AI OFFER · STRATA
AI layered onto your systems.Without touching your code.
Your information system doesn't need to be new to become intelligent. Strata adds an AI layer on top of your existing applications, even old ones, even without APIs. When the AI is unsure, an expert on our team takes over in under 30 minutes. Your systems stay untouched, and every gain is measured before it is billed.
○ AI layer off
Demo, fictitious data.
0 lines
of code changed in your applications
10 days
from diagnostic to a first prototype on your data
30 minutes
for an expert to take over a case the AI is unsure about
4 languages
French, English, Arabic and Darija
THE STRATA PRINCIPLE
An AI layer, not a rebuild project
Many AI projects stall at the same point: connecting to what already exists. You are asked to modernise first, migrate next, and AI shows up two years later. Strata flips the logic: AI sits on top of what already works.
Connected, not rewritten
We connect to your applications through their APIs, a read-only database copy, their files or their events. For software without APIs, interface agents use the screen just as a user would.
Unplugged in 24 hours
The AI layer is independent. If you switch it off, your applications work exactly as before.
Your systems readable by any AI
We expose your applications through the open MCP standard (Model Context Protocol). Your current and future AI tools can use them, with the permissions you define.
Guardrails and traceability
Access rights inherited from your systems, human approval for sensitive actions, a full log of every decision.
OUR SIGNATURE
AI handles it, our experts catch the rest
AI on its own always ends up hesitating. With Strata, every case it isn't sure about goes to an expert on our team in Casablanca, trained in your business and trilingual, who handles it in under 30 minutes. You're not buying software: you're buying completed tasks. All of them.
- 1
The AI handles the task and rates its confidence
For each task, the AI states how sure it is of its result. The threshold is set with you.
- 2
When in doubt, an expert takes over
Below the threshold, the case goes to a Clever expert who handles it in under 30 minutes, within the agreed hours, in French, English or Arabic.
- 3
Every correction improves the AI
Cases handled by our experts are used to train and test the AI. The share of cases taken over falls month after month, and you track it on your dashboard.
Why we can do it
Casablanca is one of the major French-language outsourcing hubs. Our engineers and business experts work under the same roof: AI and humans are part of the same team, with the same commitment.
THE MODEL FACTORY
The right model for every use
There is no universal model. Reading an invoice, answering a customer in Darija or predicting a failure don't need the same power or the same budget. Our router sends each task to the most efficient model, and you can switch models without rewriting anything.
Frontier models via API
For complex reasoning and writing. We work with the main providers on the market, with no exclusivity.
Open-source models on your premises
Hosted on your servers or in a cloud in your country. Your confidential data stays in.
Small specialised models
Trained or distilled from your data for a single task: faster, cheaper and able to run on modest hardware.
Predictive models
Not all AI is generative. Sales forecasting, fraud or anomaly detection, predictive maintenance: classic machine learning is often still the best choice.
We pick the model on three criteria measured on your real cases: quality, cost per task and confidentiality.
Strata Sovereign
An AI server delivered and installed on your premises, with models pre-installed, available as a monthly rental. For banks, insurers and administrations that can't use the cloud: your data never leaves your building.
Talk to an expertBUILT FOR MOROCCO AND AFRICA
Frugal AI: powerful, lean, local
No need for a multinational budget or cutting-edge infrastructure. Strata is designed for real-world conditions, in Morocco and across Africa.
Controlled costs
Small locally hosted models: less dependence on services billed in foreign currencies, and a cost per task known in advance.
Data in your country
Hosting on your servers or with a local provider, according to your rules and the applicable law.
WhatsApp and voice notes
Your customers write on WhatsApp and send voice notes: our assistants meet them there, in text and in speech.
French, English, Arabic, Darija
Assistants that understand how your customers really speak, including when they mix languages.
Locally hosted models keep working when the internet connection is slow or down.
Strata Darija
Large AI models are still weak in Darija, especially when spoken. Our teams are building their own written and spoken Darija test set to measure, select and train models on the language your customers actually speak.
THE METHOD
From first conversation to AI in production, in four steps
5 business days
Flash diagnostic
Mapping of your applications and data, a workshop with business teams, 10 costed use cases ranked by value and feasibility. You keep the roadmap, even if you go no further.
10 business days
Prototype
The top-priority use case, built on your real data, in your environment. Not a demo on sample data.
2 to 4 weeks
Shadow mode
The AI handles real cases alongside your teams, without applying anything. Its results are compared with theirs every day.
Ongoing
Go-live
Gradual switch-over as soon as the agreed quality threshold is reached, then monitoring, continuous improvement and new use cases.
PROOF FIRST
Shadow mode: we don't ask you to believe, we show you
Before any go-live, the AI works in the background on your real cases for 2 to 4 weeks. Your teams carry on as usual, and every day we measure the gap between their decisions and the AI's. The quality threshold is set with you, in writing, before kickoff. Until it is reached, nothing goes live and the subscription doesn't start.
- Quality threshold written into the contract, before kickoff
- Shared dashboard: quality, cost per task, processing time
- Every AI mistake is added to its test set
USE CASES
Where to start?
The use cases that pay off fastest, whatever your sector.
Operations
Document processing
Invoices, contracts, customer files, delivery notes: reading, checking and automatic entry into your ERP or business software.
Customer relations
Multichannel customer assistant
First-level answers on WhatsApp, email or your website, in French, English, Arabic or Darija, based on your knowledge base. An agent takes over when needed.
Data
Search across your documents
Your procedures, regulations and archives queried in plain language, with the source cited for every answer.
Data
Ask your data
"What was revenue by region in March?" Your data answers in plain language, read-only on your databases.
IT
Understand your legacy
Documentation and tests generated automatically for your older applications, before any change. Knowledge no longer leaves with departing developers.
Operations
Business agents
An end-to-end automated process, such as complaint handling, with human approval at sensitive steps.
Operations
Predictive maintenance
Anomaly detection and failure prediction from your equipment data: energy, industry, telecoms.
Customer relations
Writing and summarising
Minutes, reports, tender responses and letters, drafted from your templates and reviewed by your teams.
SECTOR PACKS
Your business, already understood
We arrive with the use cases, document templates and business rules of your sector. You don't start from a blank page.
Insurance and finance
Claims, credit files, ID and supporting document checks, customer complaints.
Energy and utilities
Field operations, meter readings, predictive equipment maintenance, customer replies.
Retail and industry
Orders, supplier invoices, delivery notes, stock and demand forecasting.
Public administrations
Citizen letters and requests, case processing, search across regulations and archives.
We connect to the most widely used software
- Sage
- Odoo
- SAP
- Microsoft Dynamics
- Oracle
- Salesforce
as well as core banking systems and custom-built business software.
OUR STANCE
Why Strata rather than a software vendor or a traditional integrator?
| Criterion | Traditional approach | Strata |
|---|---|---|
| Integration | Rebuild or proprietary connectors | Non-intrusive connectors, zero lines changed |
| AI model | Imposed by the vendor | The best one for each use, switchable at any time |
| Proof | Demo on sample data | Prototype on your data in 10 days, then shadow mode |
| When the AI gets it wrong | Your teams have to fix it | A Clever expert takes over in under 30 minutes |
| Billing | Per-user licence or man-days | On results: per task processed |
| AI costs | Variable, discovered on the invoice | Per-task cap written into the contract |
| Data | Sent to the vendor | On your premises or in your country, if you wish |
| Adoption | Delivery, then departure | Your teams trained by Clever Academy |
WHAT OTHERS DON'T DARE PROMISE
Eight commitments that take the risk off your side
01
Zero lines changed
No change to the code of your core applications. We connect alongside them, never inside.
02
No task left behind
What the AI can't do, our experts do: every uncertain case is taken over in under 30 minutes, within the agreed hours.
03
Prototype in 10 days
A working prototype on your real data, 10 business days after the diagnostic, as soon as we receive your access.
04
No go-live without proof
Shadow mode and a quality threshold written into the contract. Until it is reached, the subscription doesn't start.
05
Capped cost per task
A maximum cost per task is written into the contract. If it is exceeded, we cover the difference.
06
No lock-in
Switch models without rewriting. Code, instructions, test sets and trained models belong to you.
07
Your data stays where you decide
On your servers, in a cloud in your country or region. It is never used to train third-party models.
08
Humans stay in control
Any action that writes to your systems can require approval. Every AI decision is logged, along with what led to it.
AI READINESS TEST
Where does your company stand? Find out in 2 minutes
Eight questions to see where your company stands and which way in suits you. Nothing is saved unless you contact us.
OUR OFFERS
Four ways in, one standard: measured results
Start with your teams or with a specific case, decide on figures, pay for what gets processed. Each offer is quoted after a first 15-minute conversation.
Strata day
Your teams get hands-on with AI in one day
One day on your premises, run by Clever Academy and our engineers: training in the morning, a first use case built on your data in the afternoon.
- A working use case by the end of the day
- Your company's high-potential cases, spotted with your teams
- In Morocco: eligible for OFPPT special training contracts, subject to conditions
Flash diagnostic
Know where to start
A 5-business-day package.
- Mapping of your applications and data
- 10 costed and prioritised use cases
- 12-month roadmap
- Amount deducted from the prototype if you go ahead
Prototype and shadow mode
Proof on your data
A 10-business-day package, followed by 2 to 4 weeks of shadow mode.
- Top-priority use case on your real data
- Shadow mode on your real cases
- Quality and cost dashboard
- Go-live decision based on figures
- ONLY AT CLEVER
Pay-per-result subscription
You pay for completed tasks
A price per completed task, with a monthly minimum.
- Human safety net included: uncertain cases taken over in under 30 minutes
- Monitoring, maintenance and continuous improvement included
- Cost per task capped in the contract
- Starts only after successful shadow mode
Need help funding your project?
In Morocco and across Africa, public programmes and donors fund digital transformation. We help you identify them and put the application together. We can also be paid from a share of the savings achieved.
A multi-month AI programme? A dedicated Strata squad works for you from Casablanca, linked to your office by our Toc Toc screens. In Morocco, it comes to you.
See the offshore offerTHE CLEVER DOUBLE ENGINE
AI your teams know how to use
A tool nobody uses earns nothing. Clever is also a training organisation: every Strata project includes coaching for your teams by Clever Academy, starting with the Strata day.
Executives
Strategy, risks and investment choices: deciding with full knowledge.
Business teams
Using AI day to day, checking its answers, reporting its mistakes.
IT teams
Maintaining, monitoring and evolving the Strata layer on their own.
AI ambassadors: in each department, we train a champion who keeps adoption going after we leave.
Discover Clever AcademyCOMPLIANCE
Designed to meet the rules of every market
We work in Morocco, Africa, Canada and Europe. Every project builds in, from day one, the rules of the country where your data and users are.
Morocco
Law 09-08 on personal data protection, declarations and authorisations with the CNDP.
Africa
National data protection laws (Senegal, Côte d'Ivoire, Nigeria, Kenya and others), local hosting where required.
Canada
Quebec's Law 25 and federal PIPEDA, in French and in English.
Europe
GDPR and the AI Act: transparency obligations in force since August 2026, high-risk system obligations from December 2027.
Delivered with every project
- AI systems register
- Impact and risk assessment
- Technical documentation
- Notice to users that they are interacting with an AI
Strata is designed to make compliance easier. It does not replace advice from your lawyer or data protection officer.
Frequently asked questions about AI integration
Do we need to change our applications?
No. Strata connects alongside your applications, through their APIs, a read-only database copy, their files or their events. The code of your core applications is not changed.
Our software is old and has no API. Is it possible?
Yes, it's actually our speciality. We use the database in read-only mode, file exports, or interface agents that use the screen like a user.
Which AI models do you use?
The best fit for each task: frontier models via API, open-source models hosted on your premises, small models trained on your data or predictive models. We are not tied to any provider.
Does our data leave the company?
Only if you agree. For sensitive data, we use models hosted on your servers or in a cloud in your country. Your data is never used to train third-party models.
How long before a first result?
A 5-business-day diagnostic, then a prototype on your data in 10 business days. You see AI working on your own cases in under a month.
How do you measure quality?
On a set of real cases validated by your teams, during shadow mode and then continuously. Quality, cost per task and turnaround are visible in a shared dashboard.
What happens when the AI doesn't know?
It says so. Whenever its confidence is below the agreed threshold, the case goes to a Clever expert in Casablanca, who handles it in under 30 minutes within the agreed hours. The correction is then used to improve the AI.
How much does a project cost?
Each project is quoted after a first 15-minute conversation. The flash diagnostic then gives a precise cost for each use case, and the pay-per-result subscription sets a price per completed task, capped in the contract.
Can you help us fund the project?
Yes. We help you identify the public programmes and donors that fund digital transformation, and put the application together. In Morocco, the Strata day can be funded through OFPPT special training contracts, subject to eligibility.
What if a model changes price or disappears?
The router switches to another model without any rewrite, after validation on the same test set. And the cost per task stays capped in the contract.
Do you work in Africa outside Morocco?
Yes. We work remotely from Casablanca and on site for workshops, in French and English, with local hosting where regulations require it.
Knock knock.Where do we start?
Describe in two lines a process that wastes your time. An AI expert calls you back within 24 business hours for a first 15-minute conversation, with no commitment.