Thirty minutes with an engineer to find out what AI can really change in your organisation.
You talk directly to someone who designs AI solutions and puts them into production.
We look at your use case, your data, your tools and your constraints. If the project is worth committing to, we can then build the solution, integrate it into your information system and run it over time.
organisations equipped with our compliance software
43 days
of work recovered in one year at Straténergie thanks to a custom-built solution
Since 2019
we design and operate software in production
France & Canada
a team based in Poitiers and Montreal
the essentials
What you need to know in 30 seconds
An AI agency supports an organisation from identifying the use case through to putting an artificial intelligence solution into production.
At DPLIANCE, that scope also includes operations: monitoring, maintenance and evolution after delivery.
We build RAG document assistants, bounded AI agents, document extraction solutions and business process automation.
DPLIANCE was founded in Poitiers in 2019 and now also has a presence in Montreal. Our compliance software equips more than 500 organisations, and ADEME entrusted us with industrialising a pre-screening agent after selecting our prototype in the Dis-ADEME challenge.
Why do so many AI projects stall at the prototype stage?
Rarely because the AI model cannot produce a convincing demonstration.
The difficulties usually appear afterwards: real data more complex than expected, integration with the information system, security, running costs, governance or regulatory constraints.
It is precisely that step from demonstration to daily use that we try to prepare for from the outset.
So we test the project on representative data and define the production conditions before building an architecture that is hard to take over.
The same team designs the solution, puts it into production and then runs it. That is what allows the issues you only see in production to be handled at design time.
01
Design
The solution is designed around your processes, your data and your real environment. The prototype has to test what will actually be needed in production.
02
Go-live
We integrate the solution with your existing applications: ERP, CRM, internal tools, document repositories or line-of-business software. The goal is not to add one more isolated tool, but to place AI where it is genuinely useful.
03
Operations
After go-live, we can take on monitoring, maintenance, fixes and further development. The same team therefore stays involved in the project.
We look at the business problem, the data available, the tools involved and the main constraints. You get a first opinion on feasibility and on what needs a closer look.
2
Prototype
4 to 8 weeks
We test the solution on your data. Success criteria are agreed before development so that you can assess the result objectively.
3
Go-live
depends on the project
The solution is integrated into your information system, secured, documented and made ready to be monitored. A deeply integrated project can take several months; a more self-contained tool can be deployed much faster.
4
Operations
over time
Where you want us to, we take care of availability, maintenance and the evolution of the solution.
This path starts with a 30-minute conversation, with no commitment. You come away with a useful opinion, whether or not the rest happens with us.
Six families of solutions that we design, integrate and then operate. Each starts from the same place: a problem your teams run into, and a solution that fits into the tools already in use.
RAG & document assistants
Your teams know the information exists, but they still have to hunt for the right procedure, the right clause or the right document.
We build assistants that query an internal corpus in natural language and return answers together with their sources.
The principle matters: when an answer is going to be used in a professional context, the user must be able to check what it rests on.
Who it is for: legal, quality, HR, support, technical documentation and any organisation with a substantial document corpus.
Some processes require several operations before a person can make a decision: analysing a request, checking criteria, looking up information or drafting a reply.
We build agents whose scope of action is defined and whose operations can be logged. Sensitive steps can remain subject to human validation.
That is the approach taken for the Dis-ADEME project: preparing the assessment without taking the decision away from the caseworker.
Who it is for: back offices, support teams and departments that process requests at volume.
Shared mailboxes, qualifying requests, keeping a CRM up to date, generating documents or anonymising them: many processes are a chain of small manual tasks.
We automate those chains by integrating with your existing tools rather than forcing their replacement.
Who it is for: SMEs and mid-caps looking for quick gains without rebuilding their information system.
Control over data does not come down to a simple "hosted in Europe" promise.
You also need to look at the company operating the infrastructure, the subprocessors, the models used, any transfers and the contractual terms.
Hosting matched to the project
Our own products are hosted in Europe. For custom solutions, we can design a European or Canadian architecture, or one deployed on your own infrastructure, according to your constraints.
Models chosen to fit the need
We use European technologies such as Mistral where their performance and the project constraints allow. Other models can be considered, including in dedicated deployments, when the use case calls for it. The choice should be technical and regulatory before it is ideological.
GDPR and Law 25
Data processing, access, retention periods, traceability and any impact assessments have to be examined against the use case. Our experience in compliance tooling lets us build these questions into the architecture early.
Two locations, two regulatory environments
DPLIANCE was founded in Poitiers and now also has a presence in Montreal. This dual base lets us support European and Canadian organisations with an architecture suited to their context.
Poitiers — France & Europe
Our teams support projects governed by GDPR and design our own products from France. Hosting and suppliers are chosen according to the requirements of the project.
Montreal — Quebec & Canada
Our CTO, Florian Gadal, is based in Montreal. We also support Quebec and Canadian organisations on projects governed by Law 25 and by their own data governance constraints.
For the Dis-ADEME challenge, we designed an agent that pre-analyses funding applications. It checks certain criteria against reference data such as the SIRENE API and verifies that the items needed for assessment are present. The prototype was tested on around a hundred cases and won the challenge. ADEME then entrusted us with taking the project towards industrialisation.
The technical culture of DPLIANCE does not begin in 2019. Before co-founding the company, Hichem Ammar-Boudjelal worked as a consultant in large-account environments, while Florian Gadal worked on applications used in Quebec.
That experience of development, integration and production still shapes the way we approach Data & AI projects today.
frequently asked questions
Frequently asked questions about AI agencies
How much does an AI project cost?
The budget depends mainly on the state of the data, the complexity of the use case, the integration required and the security or compliance requirements. We price projects in stages so the prototype is validated before industrialisation is committed.
How long does an AI project take?
A prototype built on representative data generally takes four to eight weeks. Moving to production can then take anywhere from a few more weeks to several months, depending on how many systems have to be integrated and what the project requires.
What is an AI agency?
An AI agency designs and deploys artificial intelligence solutions for business problems: document assistants, AI agents, data extraction, classification or automation. At DPLIANCE, we can also take on running them after they go live.
AI provider, consultancy or agency: which should you choose?
A consultancy makes sense when what you mainly need is to define a strategy or frame a set of choices. A development provider suits a need that is already precisely defined. An agency like DPLIANCE can cover scoping, delivery and then operations. The right choice depends less on the label than on the scope you actually need.
AI agency or freelance developer: which should you choose?
A freelancer can be perfectly suited to a prototype, a targeted piece of development or an assignment that needs one specific individual expertise. For a solution meant to be integrated and run for several years, you also have to account for availability, monitoring, security, maintenance and continuity of skills. It is over that timespan that a structured team makes a difference.
Why choose a French, sovereign AI agency?
Because some projects require particular control over suppliers, the applicable jurisdiction and data transfers. That said, you should look at the actual architecture rather than the provider’s address. A French company can depend entirely on foreign services; conversely, a well-designed architecture can give precise control over the processing carried out.
What if the prototype does not work?
That is a normal possibility. The prototype exists precisely to test a hypothesis before committing to production. If the success criteria are not met, you can stop, change the approach or revisit the use case, with limited investment.
Who owns the code and the models?
The rights depend on the contract and on the components used. For bespoke development, we define before the project what is assigned, which third-party components are involved and what the reversibility terms are. It is better to make this explicit than to promise generically that "everything belongs to you".
Is our data used to train models?
Data entrusted to us as part of a project is not reused by DPLIANCE to train systems intended for other clients. The terms of use of any model providers are also taken into account when choosing the architecture.
Do you work in Quebec and Canada?
Yes. DPLIANCE has a presence in Montreal and supports projects governed by the Canadian and Quebec frameworks, in particular Law 25.
Have a use case in mind? Present it to an engineer. In 30 minutes we can already look at the problem, the data available, the main risks and how it could be tested.
Thirty minutes with an engineer who can understand your problem technically, with no salesperson in between.
A first opinion on feasibility: the possible approach, the data required and the main points to watch.
No commitment at the end of the conversation: you then decide whether the subject is worth taking further.
