9xenaiAI Workflow Automation
Connect the tools you already use so routine work runs itself.
- CRM automation
- Lead follow-ups
- Email automation
- Internal business workflows
Global AI delivery partner · Dhaka HQ
AI automation that pays for itself.
9xenai is an AI automation and consulting agency in Dhaka, Bangladesh. We help operations, finance and customer-service teams reduce manual work, improve response times and deploy AI with proper governance. Start with a fixed-scope AI opportunity assessment: a prioritised business case, a pilot plan and a measurable KPI. No vague strategy decks.
BangladeshMiddle EastEuropeAsia-PacificNorth AmericaRemote worldwide
Fixed-scope pilotsHuman-approved automationYour data, your controlSLA-backed managed servicesBangla & English delivery
Interactive operations example
- ExceptionMismatch sent to finance for approval
- SupportCustomer question answered in Bangla
- ReportMonthly sales report built and emailed
- LeadWhatsApp enquiry qualified, score 86
- InvoiceINV-4821 read and matched to its purchase order
12,512Tasks automated
44sAvg. handling time
4%Routed to people
Illustrative example with simulated data. These are not real client results or performance figures.
Business case firstOne KPI agreed before any build.
Your data, your controlYou choose where it is stored.
People approve exceptionsAutomation with human oversight.
You own the resultCode, documents and know-how.
6AI practices under one accountable team
4Decision gates: Assess, Pilot, Scale, Operate
1 dayWorking-day reply to every briefing request
2Delivery languages: Bangla and English
Start from the result you need
Select an outcome to see how we approach it and what we measure.
Remove manual handling from high-volume back-office work.
Invoice processing, reconciliation, onboarding documents and inbox triage run through automation while staff review the exceptions.
- Measured by
- Cost per transaction and hours of manual work
Drag to see what changes
One supplier invoice, from arrival to payment. Times are illustrative and vary by organisation.
Manual process
- Invoice arrives by email and waits in an inboxhours
- A clerk retypes the fields into the ERP8 min
- Staff match it to the purchase order by hand6 min
- The approver is chased by email1–2 days
- Payment is scheduled days laterdays
Several people touch every invoice.
With 9xenai
- AI reads the invoice as soon as it arrivesseconds
- Fields are extracted and validatedseconds
- Matched to the purchase order automaticallyseconds
- Approval requested only for exceptionssame day
- Payment queued with a full audit trailsame day
People see only the exceptions.
Six capabilities, one accountable team
Begin with one and add others as results are proven. Twenty-five practices are available in total. They share the same data, security model and success measures.
AI Strategy & Consulting
Find out where AI pays back, and where it doesn’t, before you commit budget.
View serviceAI Automation Solutions
Software handles the repetitive work. Your people approve the exceptions.
View serviceAI Customer Experience
Answer every customer and lead at once, in Bangla or English, from information you approve.
View serviceAI Data & Intelligence
Turn scattered files and spreadsheets into one governed view leaders can decide from.
View serviceAI Software Development
When off-the-shelf tools don’t fit, we design and build the one that does.
View serviceInnovation Lab & Training
Where new ideas become prototypes, and where your teams learn to run them.
View service
How to start
Four ways to work with us
Clear scope and a clear end point. Every engagement gets a fixed written quote after a scoping call. Indicative ranges are below, in US dollars.
Fixed fee
AI Opportunity Assessment
A 2 to 3 week assessment that identifies your three highest-ROI AI use cases, data gaps and risks, with a practical 90-day roadmap.
Book a scoping callFixed scope
AI Automation Pilot
A working pilot on one high-volume process, with baseline and post-pilot KPI comparison, a security review and a scale or no-scale recommendation.
Book a scoping callPackage
Enterprise AI Governance
Policies, AI-use inventory, risk classification, approval workflow, human-review rules, audit documentation and a monitoring plan.
Book a scoping callMonthly retainer
Managed AI Operations
We operate, monitor, improve and secure your AI assistants and automations, with a monthly performance and risk report.
See service levels$1,000+AI chatbot systems
$1,500+AI workflow automation
$2,000+AI voice agents
$500 / moManaged plans, from
Enterprise-ready from the first call
The documents and controls your security, legal and procurement teams ask for are ready before discovery starts.
NDASigned before any confidential material is shared
DPAData processing agreement available before discovery
Fixed scopeWritten acceptance criteria and change control
30 daysData return and deletion after contract end
Built on what works, not on one supplier.
We choose the right model, tool and hosting for each job, and design so any part can be swapped without rebuilding your process.
Large language modelsDocument OCR and extractionSpeech in Bangla and EnglishSearch over your own documentsWorkflow enginesAPIs and webhooksCloud, hybrid or on-premiseDashboards and BI
Size the opportunity
A planning estimate of the monthly value of automating repetitive work. The readiness review replaces it with your own process data.
Estimated monthly value
৳210,600
468 staff hours a month redirected to higher-value workAssumes about 60% of the time is automated. Illustrative only, not a forecast or a commitment.Delivery governed by decision gates
You approve each stage before the next begins, with the numbers in front of you.
Assess
Map the process, baseline the cost and agree the KPI.
Gate: business case approvedPilot
Build the smallest version and test it on real data.
Gate: KPI met in pilotScale
Integrate with core systems and train your teams.
Gate: security and operations reviewOperate
Monitor quality, cost and risk every month.
Gate: monthly service review
Global delivery
One team, delivering across regions
We are based in Bangladesh and work remote-first with organisations in international markets, with on-site sessions when they add value.
Home market
Bangladesh
On-site and remote delivery in Bangla and English, with an understanding of local systems and payment habits.
Remote-first
Middle East
Strong working-hour overlap, English delivery and project-based Arabic content and interfaces.
Remote-first
Europe
English delivery, data-location choices and documented data-protection terms for your contract.
By arrangement
Asia-Pacific and North America
Remote delivery with scheduled extended-hours sessions agreed in the proposal.
Where organisations typically start
Illustrative scenarios based on common engagements, not named clients.
Faster loan document handling
Challenge. Loan files arrive as scanned PDFs and photos that staff retype into several systems.
Approach. Document intelligence extracts the fields, checks completeness and routes exceptions to officers, with a full audit log.
Planning targetAbout 70% less manual keying
AI Automation SolutionsEvery lead answered within a minute
Challenge. Enquiries from ads, portals and WhatsApp wait hours for a reply and many go cold.
Approach. An assistant qualifies by budget, location and timeline, books viewings and updates the CRM.
Planning targetFirst reply in under 1 minute
AI Customer ExperienceFewer missed appointments
Challenge. The front desk handles bookings, changes and reminders by phone.
Approach. A Bangla and English booking assistant sends reminders and handles rescheduling automatically.
Planning targetAbout 30% fewer no-shows
AI Customer Experience
Service commitments
Measured, reported and written into the contract
Managed services run on Standard, Priority or Enterprise plans with defined severities, response targets, availability targets and a named escalation path.
30 minCritical-incident response target on the Enterprise plan, 24×7
99.9%Monthly availability target on Priority and Enterprise plans
MonthlyService review with quality, cost, risk and incident reporting
4 levelsNamed escalation from service desk to managing director
Security and governance are designed in, not added later.
Access control, audit trails and human review are part of every build.
- Choice of data location: your cloud, your servers or a local provider
- Role-based access and a full activity log
- Human approval for decisions that carry risk
- Model-neutral designs that avoid supplier lock-in
Questions leaders ask first
Straight answers before the first call.
How long does a first project take?
A typical pilot takes four to six weeks after a two to three week assessment. Larger integrations take longer, and we agree timelines in the proposal.
Do we need to move our data to the cloud?
No. We can work in your cloud, on your own servers or with a local provider. The choice is yours and is agreed before we begin.
How do you price projects?
Pilots are fixed-scope and fixed-price. Ongoing services are a monthly fee. Larger programmes are quoted by phase, with a decision gate between phases.
Will the AI make mistakes?
Sometimes, which is why we design for it. Answers come from approved sources, risky decisions go to a person, and quality is monitored every month.
Can the assistants work in Bangla?
Yes. We build and test assistants and documents in Bangla and English, including mixed-language messages.
Who owns what you build?
You do. Code, configurations, prompts and documentation are handed over, and designs avoid locking you to one supplier.
9xenai at a glance
Short factual answers about the company, its services and how engagements work.
- What is 9xenai?
- 9xenai is an AI consulting, automation and software company. It designs, builds and operates AI for organisations in Bangladesh and international markets, starting from a measurable business case.
- What services does 9xenai offer?
- AI Strategy & Consulting; AI Automation Solutions; AI Customer Experience; AI Data & Intelligence; AI Software Development; Innovation Lab & Training; WhatsApp & Chatbot Solutions; AI Content & Marketing; AI Vision & Document Intelligence; Enterprise LLM & Generative AI; Custom Model Training & Fine-tuning; Custom Enterprise AI Systems; AI Infrastructure & MLOps; Managed AI Support & Operations; AI Governance, Risk & Compliance; Agentic AI & Multi-Agent Systems; AI Voice Agents & Call Automation; AI Sales, CRM & Lead Intelligence; AI Risk, Fraud & Anomaly Detection; AI Security & Red-Teaming; Data Engineering & AI-Ready Platforms; AI for HR & Talent Operations; Bangla & Multilingual Language AI; AI Operations & Supply Chain Optimisation; AI Finance & Back-Office Automation.
- Where does 9xenai work?
- Bangladesh, the Middle East, Europe and other international markets, remote-first with on-site sessions when they add value. Assistants, documents and workshops are delivered in Bangla and English.
- How long does a first project take?
- A typical pilot takes four to six weeks after a two to three week assessment. Larger integrations take longer, and we agree timelines in the proposal.
- How do you price projects?
- Pilots are fixed-scope and fixed-price. Ongoing services are a monthly fee. Larger programmes are quoted by phase, with a decision gate between phases.
- Do we need to move our data to the cloud?
- No. We can work in your cloud, on your own servers or with a local provider. The choice is yours and is agreed before we begin.
- Who owns what you build?
- You do. Code, configurations, prompts and documentation are handed over, and designs avoid locking you to one supplier.
- Will the AI make mistakes?
- Sometimes, which is why we design for it. Answers come from approved sources, risky decisions go to a person, and quality is monitored every month.
- What service levels does 9xenai commit to?
- Managed services run on Standard, Priority or Enterprise plans with four incident severities, response-time targets (down to 30 minutes for critical incidents on the Enterprise plan), monthly availability targets of 99.5% to 99.9%, monthly service reviews and a four-level escalation path. Final values are set in the service schedule of the agreement.
Ready to see the number for your business?
Tell us which process costs you the most time. We will tell you whether AI is the right fix and what it would take.
Services
Twenty-five practices covering strategy, automation, customer experience, WhatsApp chatbots, voice agents, agentic AI, enterprise LLMs, model training, custom AI systems, MLOps, managed AI support, governance, AI security, data and data engineering, sales and CRM, finance, HR, operations, risk and fraud, Bangla language AI, content, vision AI, software and capability building.
AI Strategy & Consulting
Find out where AI pays back, and where it doesn’t, before you commit budget.
- AI readiness assessment
- Business AI roadmap
- Transformation consulting
- Implementation strategy
AI Automation Solutions
Software handles the repetitive work. Your people approve the exceptions.
- Workflow automation
- AI agents
- CRM and ERP automation
- Email and document automation
AI Customer Experience
Answer every customer and lead at once, in Bangla or English, from information you approve.
- Website AI assistants
- WhatsApp customer support
- Voice AI agents
- Lead generation and qualification
AI Data & Intelligence
Turn scattered files and spreadsheets into one governed view leaders can decide from.
- Business intelligence
- Predictive analytics
- Reporting automation
- Data processing
AI Software Development
When off-the-shelf tools don’t fit, we design and build the one that does.
- Custom AI applications
- AI-powered web platforms
- Internal business tools
- Enterprise AI integrations
Innovation Lab & Training
Where new ideas become prototypes, and where your teams learn to run them.
- Leadership and staff workshops
- Rapid prototypes
- Prompt and agent playbooks
- Pilot-to-product research
WhatsApp & Chatbot Solutions
A chatbot that answers, sells and books on WhatsApp, Messenger and your website, 24 hours a day.
- WhatsApp Business chatbot
- Order, booking and payment bots
- Messenger, Instagram and website bots
- Broadcasts and follow-up automation
AI Content & Marketing
Publish more, in Bangla and English, without hiring a bigger team.
- AI content generation
- Social media automation
- Ad creative and copy testing
- SEO and AI-search optimisation
AI Vision & Document Intelligence
Read documents, images and video automatically, and flag only what needs a person.
- OCR and document extraction
- Image classification and quality inspection
- Video and CCTV analytics
- Identity and form verification
Enterprise LLM & Generative AI
Private, governed language-model assistants that answer from your own documents, systems and policies.
- Private LLM deployment
- RAG knowledge assistants
- Enterprise copilots and agents
- Prompt engineering and evaluation
- Model selection and cost optimisation
Custom Model Training & Fine-tuning
Models trained on your data and language, when a general model is not accurate or cheap enough.
- LLM fine-tuning
- Bangla and domain language models
- Classical machine learning
- Data labelling and synthetic data
- Benchmarking and model evaluation
Custom Enterprise AI Systems
End-to-end AI products designed around your process, built to enterprise standards and integrated with your core systems.
- Bespoke AI platforms
- Multi-agent workflow systems
- Core-system integration
- White-label AI products
- Architecture and technical due diligence
AI Infrastructure & MLOps
The pipelines, hosting and monitoring that keep AI fast, reliable and affordable in production.
- Model deployment and APIs
- MLOps pipelines
- Monitoring and drift detection
- GPU and cloud cost optimisation
- Data engineering for AI
Managed AI Support & Operations
We run, monitor and improve your AI after launch, under written service levels.
- 24/7 monitoring and incident response
- Model and prompt maintenance
- Service desk and user support
- Monthly service reviews
- Dedicated AI team on retainer
AI Governance, Risk & Compliance
Put the policies, controls and audit trails in place so AI can be approved, scaled and trusted.
- AI policy and governance framework
- AI risk assessment
- Security testing and red-teaming
- Audit logs and human-in-the-loop controls
- Staff awareness and training
Agentic AI & Multi-Agent Systems
Teams of specialised AI agents that plan, use your tools and hand risky steps to a person.
- Single-task agents
- Multi-agent workflows
- Tool and API integration
- Human approval and guardrails
AI Voice Agents & Call Automation
Phone agents that answer, book and follow up, in Bangla and English, with every call logged.
- AI phone receptionist
- Appointment scheduling
- Missed-call follow-up
- Call summaries and analytics
AI Sales, CRM & Lead Intelligence
Find, score and follow up leads faster, so sales time goes to conversations that can close.
- Lead scoring and routing
- CRM enrichment and hygiene
- Follow-up sequences
- Pipeline forecasting
AI Risk, Fraud & Anomaly Detection
Spot unusual transactions, documents and behaviour early, and explain why each was flagged.
- Transaction anomaly detection
- Document fraud checks
- Risk scoring
- Alert triage
AI Security & Red-Teaming
Test your AI systems the way an attacker would, before customers or auditors do.
- LLM red-team testing
- Agent and tool permission review
- Data-flow and vendor review
- Fix plan and retest
Data Engineering & AI-Ready Platforms
Pipelines and a governed data foundation, so every AI project starts from trustworthy data.
- Data pipelines
- Warehouse and lakehouse design
- Data quality and lineage
- Vector stores and knowledge indexes
AI for HR & Talent Operations
Cut paperwork in hiring and onboarding while people keep every decision.
- CV screening support
- Interview scheduling
- Employee help assistant
- Onboarding workflows
Bangla & Multilingual Language AI
Assistants, search and document tools that handle Bangla, Banglish and English properly.
- Bangla chat and voice
- Translation and localisation
- Bangla document search
- Language quality testing
AI Operations & Supply Chain Optimisation
Forecast demand, plan stock and spot delays before they cost you money.
- Demand forecasting
- Inventory planning
- Delivery and route insight
- Supplier and order tracking
AI Finance & Back-Office Automation
Invoices, reconciliations and reports prepared automatically, with exceptions routed to your team.
- Accounts payable automation
- Bank and ledger reconciliation
- Expense and receipt processing
- Management reporting
Industries
We focus on sectors with high transaction volumes, strict controls and pressure to serve customers faster.
Banking & finance
Document-heavy workflows, strict controls and rising service expectations.
- Typical use cases
- Customer service assistants limited to approved answers
- KYC and loan document processing
- Management and risk reporting
How we deliver
A staged method that keeps risk, cost and accountability visible to your leadership throughout.
1. Assess
Two to three weeks.
- Process and data review
- Baseline cost and cycle time
- Ranked use cases and business case
Your role: sponsor, process owners, sample data.
2. Pilot
Four to six weeks.
- Working solution on real data
- Run beside the current process
- KPI comparison report
Your role: weekly reviews, user testers.
3. Scale
Two to four months.
- Integration with core systems
- Security and operations review
- Staff training and playbooks
Your role: IT, security and operations leads.
4. Operate
Ongoing.
- Monitoring and quality checks
- Monthly service report
- Continuous improvement backlog
Your role: monthly steering review.
Ways to work with us
Fixed-scope pilot
A defined outcome, timeline and price. Best for proving value before a larger commitment.
Managed service
We run, monitor and improve your automations and assistants for a monthly fee.
Dedicated team
Engineers, analysts and a delivery lead embedded with your organisation for larger programmes.
Security, data and responsible AI
Enterprise AI succeeds only when it can be trusted. These are the principles every engagement is built on.
Data location and ownership
You decide where data lives. Your data is not used to train shared models. You own the outputs and the systems we build.
Access control and audit
Role-based permissions, least-privilege access and a complete log of who and what acted on your data.
Human oversight
Decisions with financial, legal or customer risk are routed to a named person for approval.
Accuracy and testing
Assistants answer from approved sources, are tested on real cases before launch and monitored afterwards.
Supplier neutrality
We design so that models and cloud providers can be changed without rebuilding your processes.
Confidentiality
NDAs before discovery, named access for our staff and secure handling of every file you share.
What you can ask us for
Security and risk teams are welcome at the first meeting.
- Architecture and data-flow description for your use case
- Data protection and retention terms for the contract
- Responsible-AI review of the proposed automation
- Our current security and compliance status, stated plainly in the proposal
Service levels & standards
Service levels you can hold us to
Every managed engagement is backed by defined support plans, response targets, availability targets and delivery standards. Procurement, security and operations teams can review them before signing.
Support plans
Choose the plan that matches how critical the automation is to your operations. Times are in Bangladesh Standard Time (BST, UTC+6) unless the contract states otherwise.
| Plan feature | Standard | PriorityMost chosen | Enterprise |
|---|---|---|---|
| Best for | Internal tools and non-critical automation | Customer-facing assistants and finance workflows | Mission-critical, regulated or high-volume operations |
| Support hours | Business hours, Sunday to Thursday, 09:00 to 18:00 BST | Extended hours, 08:00 to 22:00 BST, five days a week | 24×7 for Severity 1; extended hours for all other severities |
| Support channels | Email and ticket portal | Email, ticket portal and chat | Email, portal, chat and a dedicated phone line |
| Availability target (managed services) | 99.5% per month | 99.9% per month | 99.9% per month, higher by agreement |
| Service review | Quarterly | Monthly | Monthly service review plus quarterly executive review |
| Named contacts | Shared service desk | Named delivery lead | Named delivery lead and engagement director |
| Recovery targets (RPO / RTO) | 24 hours / 8 hours | 8 hours / 4 hours | 4 hours / 4 hours, tested at least yearly |
How to read this. Figures are service targets. Final values, measurement method and any service credits are set in the service schedule of your agreement.
Incident severity and response targets
Every incident is classified when it is logged. Response time is how quickly a qualified engineer acknowledges and starts work.
| Severity | Definition | Standard | Priority | Enterprise |
|---|---|---|---|---|
| SEV 1Critical | Service down, data at risk or a wrong action affecting customers or money | 4 business hours | 1 hour | 30 minutes, 24×7 |
| SEV 2High | Major function degraded or a workaround is needed urgently | 8 business hours | 4 hours | 2 hours |
| SEV 3Medium | Partial issue with an acceptable workaround | 1 business day | 8 business hours | 4 business hours |
| SEV 4Low | Question, minor defect or improvement request | 2 business days | 1 business day | 1 business day |
| Update frequency (SEV 1) | Progress updates until service is restored | Every 4 hours | Every 2 hours | Every hour until restored |
After every SEV 1 and SEV 2. You receive a written root-cause report within five business days, with the actions taken and the actions planned to stop it recurring.
Escalation path
If an issue is not moving, you always know who to call next.
Level 1Service desk
Logs, classifies and acknowledges every request.
Level 2Delivery lead
Owns diagnosis and the fix for your engagement.
Level 3Engagement director
Brings in extra engineers and decision-makers.
Level 4Managing director
Executive escalation for unresolved or repeated issues.
Service standards in every engagement
These apply to fixed-scope pilots, managed services and dedicated teams alike.
Delivery
- Business case and one agreed KPI before any build
- Written acceptance criteria for every release
- Weekly working demonstrations and a short written status report
- Decision gates: you approve each stage before the next begins
- Change control: scope changes are priced and agreed in writing
Quality
- Assistants answer only from approved sources
- Tested on real cases before launch, in Bangla and English where relevant
- Ongoing sampling of conversations and outputs by a reviewer
- Documented error rate and exception volume
- Model and prompt changes tested before release
Security
- Role-based access, least privilege and a complete activity log
- Encryption of data in transit and at rest
- You choose data location: your cloud, your servers or a local provider
- Your data is not used to train shared models
- Named, NDA-bound staff only
Incident and breach handling
- Confirmed incidents affecting your data are notified to you within 24 hours
- Written incident report within five business days
- Contract terms define regulator and customer notification duties
Data protection
- Data processing agreement available before discovery
- Retention schedule agreed per engagement
- Return or deletion of your data within 30 days of contract end, with written confirmation
Hand-over and ownership
- Code, configurations, prompts and documentation are yours
- Runbooks, training and a support plan for every release
- Designs avoid locking you to one supplier
Measured and reported
You see the numbers, not a summary of effort.
- Availability against target, with every outage listed
- Incidents by severity, with response and resolution times
- Quality: accuracy samples, exception rate and hand-over rate
- Cost per transaction and model usage
- Risk log, change log and the improvement backlog
Service credits, exclusions and your responsibilities
Service credits
If monthly availability on a managed service falls below the plan target, the agreement defines credits against the monthly fee. The scale and claim process are stated in the service schedule.
Exclusions
Availability is measured excluding agreed maintenance windows, events caused by your systems or instructions, and failures of third-party services outside our control. Where agreed, we design failover to reduce that exposure.
Your responsibilities
A named sponsor and process owner, timely access to systems and sample data, review of exceptions, and notice of changes to connected systems.
Planned maintenance
Announced in advance, scheduled outside your peak hours where possible, and reported in the monthly service review.
Standards we build to
Our controls and processes are designed against recognised frameworks.
GDPR principlesISO/IEC 27001 control practicesISO/IEC 42001 AI management practicesNIST AI Risk Management FrameworkOWASP Top 10 for LLM applicationsBangladesh data-protection requirements
Alignment is not certification. We state our current certification and audit status plainly in each proposal, and we will complete your supplier security questionnaire during procurement.
Service credits
If monthly availability falls short of your plan target, credits apply against the monthly managed-service fee.
| Shortfall against plan target | Credit (% of monthly fee) |
|---|---|
| Up to 0.5 percentage points below target | 5% |
| More than 0.5 and up to 2 points below | 10% |
| More than 2 points below | 20% |
Typical scale. Credits are capped at 30% of the monthly managed-service fee and are claimed within 30 days of the monthly report. Final values are set in your service schedule.
How a support request is handled
Step 1Log and classify
Every request gets a severity when it is logged.
Step 2Acknowledge
A qualified engineer responds within the target for your plan.
Step 3Resolve
Updates at the agreed interval until service is restored.
Step 4Review
Root-cause report and preventive actions for SEV 1 and SEV 2.
Who does what
Responsibilities are written into the service schedule.
| Activity | 9xenai | Your organisation |
|---|---|---|
| Monitoring and incident response | Responsible | Informed |
| Model, prompt and workflow updates | Responsible | Approves material changes |
| Approving exceptions and risky actions | Provides the tooling | Responsible |
| Granting and revoking system access | Requests | Responsible |
| Business-rule changes | Advises and implements | Decides |
| Security fixes to systems we operate | Responsible | Informed |
| Monthly service review | Prepares and leads | Attends and signs off actions |
Change and maintenance
| Change type | Notice | Approval |
|---|---|---|
| Planned maintenance | At least 5 business days, outside your peak hours where possible | Notification |
| Workflow, prompt or content change | Agreed in the monthly review or release plan | Your process owner |
| Major model or provider change | At least 15 business days, with test results against the baseline | Your sponsor |
| Emergency security fix | As soon as practical, with notice within 24 hours after | Reviewed at the next service review |
AI quality measures we report
Targets are set from your baseline during the assessment.
| Measure | How it is measured | Reported |
|---|---|---|
| Answer accuracy | Reviewer sampling of outputs against approved sources | Monthly |
| Hand-over rate | Share of requests routed to a person | Monthly |
| Exception volume | Items needing human approval or correction | Monthly |
| Response time | Time to first reply and time to resolution | Monthly |
| Cost per transaction | Model, infrastructure and support cost divided by volume | Monthly |
| Incidents | Count by severity with response and resolution times | Monthly |
Backup, continuity and exit
- Encrypted daily backups for managed services, kept for 30 days
- Restore tested at least once a year, quarterly on the Enterprise plan
- RPO and RTO as set by your support plan
- Transition assistance and data return within 30 days of contract end
Frequently asked questions
What counts as downtime?
Time when the managed service is unavailable to your users, excluding agreed maintenance windows, events caused by your systems or instructions, and failures of third-party services outside our control.
How do I claim a service credit?
Credits are calculated from the monthly service report. You claim within 30 days of the report and credits are applied to the next invoice.
Who decides the severity of an incident?
We classify each incident when it is logged using the severity definitions, and you can ask us to raise it if the business impact is higher than it appears.
Can the targets be changed?
Yes. Higher targets, longer hours and dedicated staff can be agreed for critical operations, and are written into your service schedule.
Ready to see the number for your business?
Tell us which process costs you the most time. We will tell you whether AI is the right fix and what it would take.
Global delivery
An international AI agency, built for remote-first delivery
Based in Bangladesh, working with organisations in the Middle East, Europe and other markets. Assistants, documents and workshops are delivered in Bangla and English, and other languages are scoped per project.
Regions and working-hour overlap
Our core hours are 09:00 to 18:00 BST (UTC+6). Extended-hours sessions can be scheduled for workshops, launches and reviews.
| Region | Typical offset from BST | Overlap with core hours | Delivery model |
|---|---|---|---|
| Bangladesh | Same time zone | Full day | On-site and remote |
| Gulf and Middle East | About 2 to 3 hours behind | Most of the working day | Remote-first, on-site by arrangement |
| Europe | About 4 to 6 hours behind | Your morning | Remote-first, scheduled extended-hours sessions |
| Asia-Pacific | About 1 to 4 hours ahead | Most of the working day | Remote-first |
| North America | About 9 to 12 hours behind | Scheduled sessions | Remote-first with agreed handover windows |
Note. Offsets change with daylight-saving time. Meeting windows and support hours for your engagement are written into the proposal.
Working with international teams
Languages
Delivery in Bangla and English, including mixed-language customer messages. Arabic, Urdu, Hindi, Spanish, Dutch and Chinese content or interfaces can be scoped per project.
Data location
Your cloud, your own servers or a local provider, chosen with you before we begin. Region options are agreed in the architecture description.
Procurement
NDA before discovery, English-language contracts, a data processing agreement and supplier security questionnaires. Invoicing currency and payment terms are set in the proposal.
Governance across time zones
Weekly demonstrations, written status reports, a shared decision log and a named delivery lead who owns your engagement.
Local context
We understand Bangla-language customers, local systems and payment habits, and build to international security practice.
Hand-over
Runbooks, training and documentation so your regional teams can operate what we build.
Ready to see the number for your business?
Tell us which process costs you the most time. We will tell you whether AI is the right fix and what it would take.
Prices you can plan around
Fixed-fee builds, then an optional monthly plan to keep them running. Every range below is quoted in US dollars and turned into a written fixed quote before you commit.
- One-time builds
- $1,000 – $7,500
- Monthly plans
- $500 – $5,000+
- Billed in
- USD
Four builds, each with a clear fee
Start with one. Most clients begin with the build closest to their biggest bottleneck and add the rest later.
AI Chatbot Systems
An assistant that answers from your approved sources and hands over to a person.
- Website AI assistants
- Customer support
- Knowledge-base integration
- Lead qualification
AI Voice Agents
Phone agents that pick up, book and follow up, in Bangla and English.
- AI phone receptionists
- Appointment scheduling
- Missed-call follow-ups
- Customer inquiry handling
AI Consultancy & Implementation
Find where AI pays back, then plan and govern it before anything is built.
- AI readiness audits
- Business process analysis
- AI implementation roadmaps
- Staff training and governance
Monthly plans keep it working
Models, integrations and your own business rules change. A plan means someone is watching, fixing and improving what we built.
Essential support
$500
per monthKeeps one live system healthy and reported on.
Growth automation
$1,000
per monthSupport plus steady improvements to your automations.
Advanced AI operations
$2,000
per monthFor several connected systems that the business depends on.
Enterprise AI partnership
$3,000
to $5,000+ per monthA standing AI team for critical operations, shaped to your needs.
In every plan
MonitoringMaintenanceOptimisationReportingDefined support allowance
Charged separately
Substantial new developmentUsage costs
Response targets, availability and service credits for managed plans are set out in service levels and standards.
Estimate your first year
Pick the builds you are considering and a monthly plan. This is a planning range, not a quote.
Packages tuned for two sectors first
Specialising makes scoping faster and prices more predictable. We focus first on real estate and marketing agencies, and take other sectors case by case.
Real estate
Enquiries arrive from portals, ads, WhatsApp and calls. The value is in answering all of them fast.
- CRM automation and lead follow-ups
- Chatbot that qualifies buyers and tenants
- Voice agent for calls and missed-call follow-ups
Marketing agencies
Repetitive reporting, approvals and client onboarding eat the hours that should go to strategy.
- Workflow and email automation for client operations
- Chatbot for lead capture on client sites
- Readiness audit and staff training
We help businesses reduce operational costs, automate repetitive work, improve lead conversion and scale with AI-powered systems.
We sell business outcomes, not chatbots. Every build starts from a number we can measure, so you can see whether it paid off.
The APEX method
How our consulting engagements run, from first audit to measured results.
Analyze
Audit your business, operations, challenges and opportunities.
Plan
Create a customised AI transformation roadmap.
Engineer
Build AI agents, automation systems, integrations and intelligent workflows.
eXpand
Measure results, optimise performance and scale what works.
Pricing questions
Are these fixed prices?
Each range is indicative. After a scoping call we send a fixed written quote with scope, timeline and acceptance criteria. The price changes only through agreed change control.
What is in a monthly plan?
Monitoring, maintenance, optimisation, reporting and a defined support allowance. The allowance and response targets are written into your service schedule.
What costs extra?
Substantial new development and usage costs, such as model, API and telephony charges. We estimate usage in the proposal so there are no surprises.
Do I have to take a monthly plan?
No. You can buy a build on its own and own the result. We recommend a plan for anything your customers rely on, because models and integrations keep changing.
Which currency do you bill in?
US dollars.
Get a fixed quote for your build
Tell us which process costs you the most time. After a free 30-minute call you get a written quote.
Enterprise delivery you can defend in a steering committee
How we plan, govern, test and support AI work, so your leadership, security and procurement teams can approve it with confidence.
What enterprise buyers ask for, and where to find it
Each question below has a page with the detail, so your reviewers do not have to wait for a call.
Is it secure?
Access control, encryption, logging, vulnerability handling and AI-specific controls, with the evidence you can request.
Trust centreCan procurement approve us?
NDA, data processing agreement, MSA, service schedule, questionnaires and a defined exit path.
Procurement packWho handles your data?
Roles, sub-processors, breach notice within 24 hours, transfers and deletion within 30 days.
Data processing termsWhat happens after go-live?
Severity-based response targets, monthly reporting, credits and a named escalation path.
Service levelsHow is AI risk controlled?
Human approval for risky actions, tested answers, change control and an AI use policy.
Security & governanceWhat will it cost?
Indicative ranges in US dollars and a fixed written quote after the assessment.
PricingReference architecture
Every solution follows the same pattern, so your architects can review it once and recognise it in each project.
How an engagement is governed
Decisions stay with your leaders. These forums keep progress, risk and spend visible.
| Forum | Who attends | When | What it decides |
|---|---|---|---|
| Steering committee | Your sponsor and leads, our engagement lead | Monthly | Priorities, budget, scope changes and risks needing leadership |
| Weekly working demo | Process owners, our delivery team | Weekly | Whether what was built works on real cases |
| Risk and decision log | Project leads on both sides | Updated continuously | Open risks, owners, dates and decisions taken |
| Stage gate review | Sponsor and our engagement lead | End of each stage | Whether to proceed, change course or stop |
| Monthly service review | Service owner and our operations lead | Monthly after go-live | Quality, cost, incidents and the improvement backlog |
Who is on the engagement
Named roles with clear responsibilities, on both sides.
From 9xenai
- Engagement lead. Accountable for scope, plan and reporting.
- Solution architect. Owns the design and data flows.
- AI and automation engineers. Build and test.
- Evaluation and QA lead. Owns test sets and release checks.
- Security and governance lead. Owns controls and evidence.
From your organisation
- Executive sponsor. Owns the business case and approves each gate.
- Process owner. Defines the rules and accepts the results.
- IT and security contact. Grants access and reviews the design.
- Data owner. Approves data sources and sharing.
How we test before anything goes live
| Check | What is tested | Evidence you receive |
|---|---|---|
| Baseline measurement | Time, cost and error rate of the current process | Baseline sheet signed off by the process owner |
| Evaluation set | Answers and actions against approved sources and known cases | Accuracy report before each release |
| Parallel run | The automation beside the manual process on live work | Side-by-side results and exception list |
| User acceptance | Your staff try real scenarios against written acceptance criteria | Signed acceptance record |
| Security and abuse tests | Prompt injection, data leakage and permission limits | Findings ranked by severity, with retest results |
| Rollback plan | How to switch back to the manual process quickly | Tested rollback steps in the runbook |
What you receive at hand-over
You own the result and can run it without us.
- Business case and KPI baseline
- Architecture and data-flow description
- Evaluation report and acceptance record
- Runbook, rollback steps and monitoring guide
- Source code, configurations and prompts
- Staff training and admin guide
- Risk log and the improvement backlog
Typical risks and how we control them
| Risk | Typical cause | Control |
|---|---|---|
| The pilot works but never scales | No owner or measured business case | KPI and sponsor agreed before build, stage gate before scale |
| Wrong or invented answers | Model answers beyond approved content | Answers limited to approved sources, evaluation set, hand-over to a person |
| Data exposure | Excess access or an unsuitable model provider | Least privilege, named providers, no training on your data, region of your choice |
| Supplier lock-in | Design tied to one model or cloud | Swappable model layer, documented hand-over, data return within 30 days |
| Quality drifts after launch | Changing data, prompts or provider models | Monthly quality sampling, change control and regression tests |
| Scope and cost creep | Unclear requirements | Fixed scope, written acceptance criteria and change control |
Alignment is not certification. Our practices are designed against GDPR principles, ISO/IEC 27001 and 42001 control practices, the NIST AI Risk Management Framework and the OWASP Top 10 for LLM applications. See the trust centre for our exact status.
Enterprise questions
Can we use our own contract template?
Yes. We can work from your master agreement or ours, and agree the statement of work and service schedule alongside it.
Will you work with our security team directly?
Yes. We schedule an architecture review with your security team before build, and answer questionnaires within five business days of request.
Can you provide client references?
Reference availability depends on each client's permission. Where we cannot name clients, we offer a small fixed-scope pilot so you can judge the work first.
Who owns the intellectual property?
You own the deliverables defined as yours in your agreement. Code, configurations, prompts and documentation are handed over.
What if a model provider changes or is withdrawn?
Designs separate the model layer from your process, so a provider can be changed without rebuilding. Major changes are tested against your baseline and notified in advance.
Can the service run inside our own environment?
Yes. We can work in your cloud, on your servers or with a local provider. The region and any model provider are named in the architecture description before work starts.
Bring your security and procurement teams to the first call
Send your questionnaire or NDA. We reply within one working day.
Products
Repeatable solutions grown from client delivery. Each can be deployed as a managed service.
9xenai Reply
WhatsApp and website assistant with approved answers, lead qualification and human hand-over.
Customer experience9xenai Docs
Reads invoices, contracts and forms, extracts the fields and posts them into your finance or CRM system.
Automation9xenai Insight
One governed view of sales, stock and finance, with forecasts and plain-language questions.
Data & intelligence9xenai Voice
Phone agent for bookings and enquiries in Bangla and English, with every call logged.
Customer experienceRequest an executive briefing
Tell us about the process or outcome you want to improve. A senior consultant will reply within one working day.
Request prepared
About 9xenai
We help organisations do far more with the people and systems they already have. AI is our tool. Measurable business results are our product.
Why the name
9x stands for leverage. The aim of every project is to multiply what a team can achieve, not to replace the team. Work that is repetitive, slow or error-prone moves to software. People keep the judgement, the relationships and the decisions.
We work from Bangladesh and deliver to clients in Bangladesh and international markets, in Bangla and English.
What we stand for
Outcomes over hype
We sell results you can measure. If we cannot define the KPI, we do not start building.
People stay in control
Automation handles the routine. Anything risky goes to a named person for approval.
Built to be owned
You receive the code, the documents and the know-how, and your team is trained to run it.
Local context, global standards
We understand Bangla-language customers, local systems and payment habits, and build to international security practice.
How we work with you
Bilingual delivery
Workshops, assistants and documents in Bangla, English or both.
Weekly demonstrations
You see working software every week, not a report at the end.
Remote-first
Distributed delivery, with on-site sessions when they add value.
Documented hand-over
Runbooks, training and a support plan for every release.
Case studies
Six scenarios that show how organisations in different sectors typically begin.
These are illustrative scenarios based on common engagements. Targets are planning goals, not guaranteed results or results for a named client.
Faster loan document handling
Challenge. Loan files arrive as scanned PDFs and photos that staff retype into several systems.
Approach. Document intelligence extracts the fields, checks completeness and routes exceptions to officers, with a full audit log.
Planning targetAbout 70% less manual keying
AI Automation SolutionsEvery lead answered within a minute
Challenge. Enquiries from ads, portals and WhatsApp wait hours for a reply and many go cold.
Approach. An assistant qualifies by budget, location and timeline, books viewings and updates the CRM.
Planning targetFirst reply in under 1 minute
AI Customer ExperienceFewer missed appointments
Challenge. The front desk handles bookings, changes and reminders by phone.
Approach. A Bangla and English booking assistant sends reminders and handles rescheduling automatically.
Planning targetAbout 30% fewer no-shows
AI Customer ExperienceBuyer documents without the chase
Challenge. Orders, packing lists and compliance forms are re-entered in several systems for every shipment.
Approach. Automation reads buyer documents, fills order and shipping records and flags any mismatch.
Planning targetAbout 60% less admin per shipment
AI Automation SolutionsSupport that scales with sales
Challenge. Enquiries spike during campaigns and response times slip.
Approach. A WhatsApp assistant handles order status, returns and common questions, with hand-over to staff.
Planning target3× the enquiries from the same team
AI Customer ExperienceOne view of delivery performance
Challenge. Hub performance sits in separate spreadsheets that are merged by hand each week.
Approach. A unified dashboard with late-delivery alerts and route analytics refreshes automatically.
Planning targetDaily reporting in minutes, not days
AI Data & IntelligenceInsights
Practical guidance for leaders planning their first AI projects.
How to choose your first AI use case5 min read
The best first project is small, frequent and measurable. It should prove value quickly and teach your team how to work with AI.
- Pick a process that repeats at least weekly
- Choose one with a clear cost or time baseline
- Avoid processes with no accessible data or named owner
- Define the KPI and a stop rule before building
What a WhatsApp assistant should and should not handle4 min read
A good assistant is clear about its limits. Let it answer routine questions fast, and hand over everything that needs judgement.
- Handle: opening hours, prices, order status, bookings and lead capture
- Hand over: complaints, refunds, legal questions and any request to speak to a person
- Answer only from approved content
- Log every conversation and review samples weekly
What an AI pilot must prove before you scale5 min read
A pilot is an experiment with a decision at the end. Agree the pass mark before it starts.
- The KPI beat the baseline on real data
- The error rate and exception volume are known
- Users accept the tool and use it
- Security and data owners have signed off
- The running cost per transaction is understood
A data readiness checklist for AI projects4 min read
Most delays come from data, not models. Ask these questions in the first week.
- Where does the data live and who owns it?
- Is it digital, structured enough and up to date?
- Can we reach it through an export or an API?
- Does it contain personal data that needs extra controls?
- Who corrects errors when we find them?
Privacy Policy
How 9xenai handles personal information. Last updated: November 2026.
What we collect
The details you enter in the contact form (name, work email, organisation, role, region, area of interest, timeline and your message). The site itself stores only your language and light or dark theme choice in your browser.
How we use it
To reply to your request, prepare a proposal and manage the business relationship. We do not sell personal information.
Sharing
Only with service providers that help us run the website, email and scheduling, under confidentiality terms, or where the law requires it.
Retention and your rights
We keep enquiry details only as long as needed for the purpose above. You can ask us to access, correct or delete your information at any time using the contact details on this site.
Third-party services
The site loads fonts from Google Fonts. Calendar or form providers, if used, are named on the page where they appear.
Client data
Data processed during an engagement is governed by the signed agreement and data processing agreement, which are available before discovery.
Terms of Service
Terms for using this website. Engagements are governed by a separate signed agreement. Last updated: November 2026.
Use of the site
The content is provided for general information. You agree not to misuse the site or attempt to disrupt it.
No guarantee of results
Case studies, calculators and targets on this site are illustrative planning aids, not forecasts, guarantees or results for a named client.
Engagements
Scope, fees, timelines, service levels, ownership of deliverables and liability are set in the proposal and service agreement for each engagement.
Intellectual property
Site content and the 9xenai name belong to 9xenai. You own the deliverables defined as yours in your agreement.
Changes
We may update these terms. The date at the top shows the latest version.
AI use policy
How we use AI in our own work and in client delivery. Last updated: November 2026.
Human oversight
Decisions with financial, legal or customer risk are routed to a named person for approval.
Your data
Your data is not used to train shared models. You choose where it is stored, agreed before work begins.
Accuracy and limits
AI can make mistakes. Assistants answer from approved sources, are tested before launch and are sampled and monitored afterwards.
Transparency
Customer-facing assistants identify themselves as automated and offer a hand-over to a person.
Frameworks
Our practices are designed against GDPR principles, ISO/IEC 42001 and the NIST AI Risk Management Framework. This is alignment, not certification.
Trust centre
Security and trust, documented before you sign
Security, legal and procurement teams can review our controls, data-handling terms and AI governance practices before the first workshop. Where we align to a framework but are not certified, we say so.
Security controls
What we commit to in every engagement, and the evidence your team can ask for.
| Area | What we commit to | Evidence on request |
|---|---|---|
| Access control | Role-based access and least privilege. Multi-factor authentication on all staff accounts that can reach client systems. Access reviewed every quarter and removed at contract end. | Access matrix for your engagement |
| Encryption | TLS 1.2 or higher in transit. Encryption at rest for stored client data and backups. | Architecture and data-flow description |
| Logging and audit | Activity logs for access to client data and for every automated action, kept for the period set in the contract. | Redacted sample log extract |
| Secure development | Peer review of every change, automated dependency and secret scanning, and separate development, test and production environments. | Release and change log |
| Vulnerability management | Critical vulnerabilities in systems we operate are triaged within 24 hours and fixed or mitigated within 7 days. High-severity issues within 30 days. | Patch summary in the monthly review |
| Backup and recovery | Encrypted backups for managed services. Restore tested at least once a year. RPO and RTO follow your support plan. | Restore-test summary |
| People | NDA and confidentiality terms for every employee and contractor, security awareness training on joining and every year, named access only. | Training and NDA confirmation |
| Third parties | Model, cloud and tool providers are listed as sub-processors. You are told about changes before they take effect. | Sub-processor list |
| Data location | You choose the region: your cloud, your servers or a local provider. | Region options in the architecture description |
AI-specific controls
Risks that conventional security checklists miss.
| Risk | Control |
|---|---|
| Wrong or invented answers | Answers limited to approved sources, with citations where relevant. An evaluation set runs before every release. |
| Prompt injection and data leakage | Input and output filtering, tool permissions limited per task, and testing against the OWASP Top 10 for LLM applications. |
| Risky automated actions | Human approval for actions with financial, legal or customer impact, plus spending and action limits. |
| Model and prompt changes | Every change is tested against the baseline set, logged and released through change control. |
| Training on your data | Not permitted. Third-party model APIs are configured for no training and the shortest retention available. |
| Bias and unfair outcomes | Reviewer sampling, documented limits and a route to a named person for any contested outcome. |
Compliance status
Stated plainly, so you do not have to guess.
| Framework | Our position |
|---|---|
| GDPR principles | Designed to. A data processing agreement is available before discovery. |
| ISO/IEC 27001 | Control practices aligned. Not certified. |
| ISO/IEC 42001 | AI management practices aligned. Not certified. |
| NIST AI Risk Management Framework | Used to structure risk reviews of each AI system. |
| SOC 2 | Not certified. Your security team can request control evidence and we complete questionnaires. |
| OWASP Top 10 for LLM applications | Used in design reviews and testing. |
Alignment is not certification. If this changes, this page will say so and name the auditor and scope.
Documents available under NDA
Ask for any of these during evaluation.
- Security questionnaire response (CAIQ, SIG-lite or your own template)
- Architecture and data-flow description for your use case
- Sub-processor list and model-provider details
- Business continuity and recovery summary
- Responsible-AI review of the proposed automation
- Sample service schedule with severities, credits and escalation path
Responsible disclosure
If you believe you have found a vulnerability in our website or systems, please tell us.
Use the contact in our security.txt file. We acknowledge reports within two business days, keep you updated, and do not pursue good-faith research that avoids harming users or data.
Frequently asked questions
Are you ISO 27001 or SOC 2 certified?
We state our certification status plainly in each proposal. Our practices are aligned to ISO/IEC 27001 control practices and we complete your supplier security questionnaire, but alignment is not certification.
Where is our data processed and stored?
Where you choose: your cloud, your own servers or a local provider. The region and any third-party model provider are written into the architecture description before work begins.
Is our data used to train AI models?
No. Your data is not used to train shared models. Where a third-party model API is used we configure it for no training and the shortest retention available, and we name the provider to you.
How do I report a security issue?
Email the security contact listed in our security.txt file. We acknowledge reports within two business days and treat good-faith research as welcome.
Questions from your security, legal or procurement team?
Send us your questionnaire or NDA. We reply within one working day.
Confidentiality
NDA signed before discovery begins
We sign a mutual non-disclosure agreement before any confidential material is shared, and we treat client data in AI projects with the same care as the documents themselves.
Standard mutual NDA: key terms
A summary of our usual position. We are happy to sign yours.
| Term | Standard position |
|---|---|
| Type | Mutual. Both parties protect each other’s information. |
| Timing | Signed before any confidential material is shared. We return our signed copy within one working day. |
| What is protected | Business, technical, financial and customer information in any form, including data samples and system access. |
| Exclusions | Information that is public, already known, independently developed or lawfully received from a third party. |
| Permitted use | Only to evaluate and deliver the agreed engagement. |
| Who can see it | Named employees and contractors who need it and are bound by equal terms. |
| Duration | Typically three years from disclosure. Trade secrets and personal data stay protected for as long as they remain so. |
| Return or deletion | Within 30 days of your request or the end of the engagement, with written confirmation. |
| Compelled disclosure | Notice to you first, where the law allows. |
| Governing law | Agreed per engagement. |
Summary only. The signed agreement governs. Have your own counsel review any agreement before you sign.
How we protect confidential material in AI projects
Data minimisation
We ask for the smallest sample that proves the use case and mask or anonymise personal data wherever possible.
Approved tools only
Client material is processed only in environments agreed with you. Third-party model providers are named and configured for no training.
Separate workspaces
Each client has an isolated workspace. Access is named, logged and removed at the end of the engagement.
Prompts and logs
Prompts, outputs and logs that contain your information are confidential and deleted under the agreed retention schedule.
How it works
Step 1Request
Ask for our template or send yours.
Step 2Review
We review and reply within one working day.
Step 3Sign
Electronic signature by both parties.
Step 4Share
Discovery begins and material is shared.
Frequently asked questions
Will you sign our NDA?
Yes. We can sign your standard NDA or use our mutual template. Send it with your first message and we will review it within one working day.
Do we need an NDA before the first call?
Not for a general introduction. We ask for an NDA before you share any confidential documents, data samples or system access.
Can our confidential data be sent to a public AI tool?
No. Client material is processed only in approved environments agreed with you. Consumer chat tools are not used for client data.
What happens if a team member leaves?
Every employee and contractor is bound by confidentiality terms that continue after they leave, and their access is removed the same day.
Questions from your security, legal or procurement team?
Send us your questionnaire or NDA. We reply within one working day.
Data protection
Data processing terms you can review before you sign
A data processing agreement is available before discovery begins. This page summarises our standard position on roles, sub-processors, breaches, transfers and deletion.
Processing summary
Final values are set in the data processing schedule of your agreement.
| Item | Standard position |
|---|---|
| Roles | You are the controller. 9xenai is the processor. |
| Instructions | Personal data is processed only on your documented instructions. |
| Data and purposes | Categories of data, data subjects and purposes are listed per engagement in the data processing schedule. |
| Duration and retention | The term of the engagement, plus any retention period you specify. |
| Sub-processors | Listed in the agreement. You are told about changes in advance and may object. |
| Security measures | As described in the Trust centre and the agreement. |
| Personal data breach | Notified to you within 24 hours of confirmation, with a written report within five business days. |
| Data subject requests | We help you respond promptly, usually within five business days of your request. |
| International transfers | Region chosen by you. Where a transfer mechanism is required, it is documented in the agreement. |
| Audit and evidence | Questionnaires and evidence on request. Audits by arrangement as set out in the agreement. |
| Return and deletion | Returned in a standard format and deleted within 30 days of contract end, with written confirmation. |
Not legal advice. The signed agreement governs. The applicable data-protection laws are named in it.
Frequently asked questions
Who is the controller and who is the processor?
You are the controller of your data. 9xenai acts as processor and handles personal data only on your documented instructions.
How quickly will you tell us about a breach?
Within 24 hours of confirming an incident that affects your data, followed by a written report within five business days.
Can we audit you?
You can request security evidence and questionnaires during the contract, and audits by arrangement where the agreement provides for them.
What happens to our data when the contract ends?
It is returned in a standard format and deleted within 30 days of contract end, with written confirmation.
Questions from your security, legal or procurement team?
Send us your questionnaire or NDA. We reply within one working day.
Enterprise procurement
Everything procurement needs, in one pack
Standard documents, clear contract structure and a defined exit path, so your vendor review does not become the slowest part of the project.
Documents we provide
| Document | What it covers | When |
|---|---|---|
| Mutual NDA | Confidentiality of material shared during evaluation and delivery | Before discovery |
| Data processing agreement | Roles, instructions, sub-processors, breach notice, deletion | Before discovery |
| Master services agreement and statement of work | Scope, deliverables, acceptance criteria, fees, change control | With the proposal |
| Service schedule | Support plan, severities, response and availability targets, credits, escalation | With the proposal |
| Security questionnaire response | Answers on your template, CAIQ or SIG-lite | Within five business days of request |
| Architecture and data-flow description | Components, data location, model providers, access paths | With the proposal |
| Sub-processor list | Providers that may process your data, with purpose and region | With the proposal |
| Responsible-AI review | Risks, controls and human-approval points for the proposed automation | During the assessment |
| Vendor onboarding details | Company registration and tax information for your supplier forms | On request |
Typical path to signature
Week 1Intro and NDA
Short call, NDA signed, scope outline.
Weeks 1 to 2Assessment
Process review and security questionnaire.
Weeks 2 to 3Proposal
Scope, KPI, fixed pilot price, service schedule.
Weeks 3 to 4Contract
MSA, SoW and DPA review and signature.
Exit and continuity
You should never be locked in.
You own the result
Code, configurations, prompts and documentation are handed over.
Transition assistance
Runbooks, knowledge transfer and a handover period agreed in the contract.
Data return
Your data is returned in a standard format and deleted within 30 days of contract end.
Supplier neutrality
Designs allow models and cloud providers to be changed without rebuilding your process.
Frequently asked questions
How long does vendor onboarding take?
Most organisations complete NDA, questionnaire and contract review within two to four weeks. We respond to each request within one working day.
Can we use our own contract template?
Yes. We can work from your master agreement or use ours, and we agree the statement of work and service schedule alongside it.
What if we want to leave?
You own the code, configurations, prompts and documentation. We provide transition assistance and return your data within 30 days of contract end.
Do you work with smaller organisations?
Yes. The same documents and fixed-scope pilots apply, scaled to the size of the engagement.
Questions from your security, legal or procurement team?
Send us your questionnaire or NDA. We reply within one working day.
Services / Practice 19 of 25
AI Risk, Fraud & Anomaly Detection
Spot unusual transactions, documents and behaviour early, and explain why each was flagged.
Short answer
AI Risk, Fraud & Anomaly Detection is a 9xenai AI practice. Spot unusual transactions, documents and behaviour early, and explain why each was flagged. Built for finance, compliance and operations teams reviewing high transaction volumes. Typical target: Daily target review cadence instead of month-end sampling.
Built for: Finance, compliance and operations teams reviewing high transaction volumes.
Target outcomeDailytarget review cadence instead of month-end samplingTargets are agreed with you in the assessment and measured against your own baseline.
What is included
Transaction anomaly detection
Flag outliers in payments, claims, expenses and orders.
Document fraud checks
Compare invoices, IDs and statements for inconsistencies.
Risk scoring
Rank cases for review with a plain-language reason for each.
Alert triage
Group alerts and cut false positives so analysts review what matters.
Works with
ERP and finance systemsPayment dataCase management toolsData warehouse
See a typical job run
Illustrative figures show the kind of change we aim for. Your baseline sets the real numbers.
- Transactions and documents are collected
- Models score each item
- Flags come with a clear reason
- Analysts confirm or clear each case and the model learns
Share of items reviewed (%)5Typical today: 5
False alerts per 10040Typical today: 40
How an engagement unfolds
Week 1
Define what counts as risk and gather labelled historical cases.
Weeks 2–4
Run the model beside your current checks and compare findings.
Month 2 onward
Move to daily review with analyst feedback built in.
Questions we are asked
Can it make decisions on its own?
It ranks and explains. A named reviewer decides on any case that affects a customer or a payment.
Do you need a lot of historical data?
It depends on the use case. We say in the assessment whether your data is enough and what to collect if it is not.
Other services
AI Strategy & ConsultingAI Automation SolutionsAI Customer ExperienceAI Data & IntelligenceAI Software DevelopmentInnovation Lab & TrainingWhatsApp & Chatbot SolutionsAI Content & MarketingAI Vision & Document IntelligenceEnterprise LLM & Generative AICustom Model Training & Fine-tuningCustom Enterprise AI SystemsAI Infrastructure & MLOpsManaged AI Support & OperationsAI Governance, Risk & ComplianceAgentic AI & Multi-Agent SystemsAI Voice Agents & Call AutomationAI Sales, CRM & Lead IntelligenceAI Security & Red-TeamingData Engineering & AI-Ready PlatformsAI for HR & Talent OperationsBangla & Multilingual Language AIAI Operations & Supply Chain OptimisationAI Finance & Back-Office Automation
Governance included
Every AI Risk, Fraud & Anomaly Detection engagement carries the same safeguards.
- Business case and KPI agreed before build
- Role-based access and a complete activity log
- Human approval for decisions that carry risk
- Documentation and hand-over so you own the result
Service commitments
Backed by service levels
Managed engagements include a support plan with severity-based response targets, availability targets, monthly reporting and a named escalation path.
30 minCritical response target (Enterprise plan)
99.9%Monthly availability target (Priority and Enterprise)
MonthlyQuality, cost and risk reporting
Ready to see the number for your business?
Tell us which process costs you the most time. We will tell you whether AI is the right fix and what it would take.