PRODUCT
Digital Human Custom B2B
This is not a product you switch on. It is a Digital Human built around the company that will use it: its processes, its documents, its systems and the way it speaks to its customers. By the end of training it knows what to do and how to do it, within the perimeter the company has set.
Request an analysis of your business modelWHAT CUSTOM MEANS HERE
Five things built around the company, not configured
The difference between a custom Digital Human and a generic assistant is not the number of features. It is that each of these five elements is drawn from the client company and does not exist before it.
| The knowledge | Internal procedures, price lists, standard contracts, request history, the answers the company already gives today. The Digital Human does not learn from the generic web: it learns from what the company knows and how the company says it. |
| The systems | Management software, CRM, ERP, mail, calendar, warehouse. Connections are built on the systems the company actually uses, including bespoke ones, because value appears when the Digital Human executes and not only when it answers. |
| The voice | Register, terminology, level of formality, what can be promised and what cannot. A customer writing to the company must recognise the company, not a service. |
| The boundaries | What it may do alone, what it must submit to a person, what it must never do. The perimeter is decided by the company and written down before the Digital Human enters service. |
| The machine | Every client company runs on a dedicated server machine. The company's data stays there: it is not mixed with other companies' data and does not feed the training of third-party models. |
INSIDE THE TRAINING
What happens during the training months
The five-phase adoption path is described on the home page. Here is the part that usually goes untold: what concretely goes into the Digital Human during training, and who puts it there. The work is carried out by the AI engineering group at AI.ANIMA together with the people in the company who run those processes every day.
- Process mappingWe start from the business model and isolate the high-volume repetitive activities: the ones where people's time is worth less than it costs. We do not automate everything, we automate what is worth automating.
- Knowledge ingestionDocuments, procedures, catalogues and historical correspondence are structured and made queryable. The contradictions that surface, and they always do, are reported back to the company: often that is the project's first benefit.
- Building the connectionsEvery action the Digital Human will have to perform is linked to the system that executes it, and tested on non-production data until it is reliable.
- Calibrating the voiceWe work on real answers: we compare them with what the company's best person would say and correct them until they match.
- Writing the boundariesIrreversible actions are defined, along with those requiring human confirmation, and we verify that the Digital Human really does stop where it must stop.
FROM TESTING TO SERVICE
How the move from trial to real service happens
The move is not a switch. It happens in three stages, and at each one the company can stop, correct or step back with no consequence for customer service.
01
Isolated trial
The Digital Human works in a separate environment on non-production data. No customer sees it. Answer accuracy and action success rate are measured.
02
Shadowing
The Digital Human receives real requests and drafts the answer, but a person validates it before it is sent. Every human correction becomes training material.
03
In service
The Digital Human answers and acts autonomously within the agreed perimeter, while irreversible actions remain subject to confirmation. Monitoring continues and the perimeter widens only when the company decides.
QUESTIONS AND ANSWERS
The questions we are asked before signing
Where does my company's data end up?
On a server machine dedicated to your company, in Europe, on OVHcloud infrastructure, where AI.ANIMA is an Advanced Partner. The data stays safe inside the dedicated machine: it is not shared with other client companies, it is not used to train third-party models and it does not travel to outside services to be processed. Credentials to your systems remain yours and can be revoked at any time.
How long does training take and what do you need from us?
Training usually takes around three months, but that is an estimate and varies with process complexity and the quality of existing documentation. We need three things from you: the people who know the processes, access to the systems to be connected, and a clear decision on what the Digital Human may do alone. The time asked of your staff is concentrated in the first weeks and then drops.
Who runs the project and how do they work with us?
The AI engineering group at AI.ANIMA. The path starts with a meeting where we analyse the business model and identify the activities where a Digital Human produces a measurable return. From there we move to training, then testing, then service, and finally to continuous improvement that does not stop at handover. We do not drop a product and walk away.
How do I know it works before exposing it to customers?
Because you do not expose it until you have seen the numbers. During the isolated trial and shadowing stages the Digital Human works without any customer seeing it, and you measure how many answers are correct and how many actions succeed. Going into service is your decision, taken on data you have seen, not on a promise from us.
If our processes change, does the Digital Human become obsolete?
No, because training does not end at handover. Knowledge can be updated and what the company teaches stays and accumulates. When a price list, a procedure or a system changes, the affected part is updated without redoing the work.
What is left with us if one day we end the relationship?
Your data and your documentation are yours and are returned in a readable format. You revoke the credentials of the connected systems yourself. This is a point we put in writing in the contract, not a courtesy.
INVESTMENT
The price is built on the work, not on a list
Every custom Digital Human is a different project, because the processes to map, the systems to connect and the volume to sustain are different. That is why the price comes from a quotation drawn up after the business model analysis, not from a three-column table. The initial analysis is free of charge.
Request a quotationPRICE STRUCTURE
ONE-OFF
Upfront cost
Covers the business model analysis, process mapping, knowledge ingestion, building the connections to your systems and the whole path through to service. It is quoted on the work actually foreseen and agreed before the start, not billed at the end.
MONTHLY
Recurring fee
Covers the dedicated server machine, running the service, monitoring and continuous improvement. It is set against the agreed perimeter and the volume of work sustained. It changes only if the perimeter or the volume changes, and only by agreement.