The conventional consultant
You get a report, recommendations and slides. The code, the integration and the move to production remain your problem.
Our delivery model
Forward-deployed engineering puts AI to work inside your systems
Most AI projects die somewhere between the demo and the working day. Our answer is the model Palantir invented. An engineer embeds in your real environment (your ERP, your data, your rules) and stays until you have your first measured results in your own business. Then the engineer trains your team to take over.
Inside your systems
We work in your tools, on your production data, from the first week.
Accountable for results
The engineer answers for what runs in production.
Autonomy handed over
You keep the written scope, a trained team and control of oversight.
At Swiss scale
The giants' model, sized for companies of 20 to 250 employees.
In brief
What is forward-deployed engineering, and how does Djtal apply it?
Forward-deployed engineering (FDE) is a delivery model that embeds an engineer in the client's systems, data and governance to put AI into production. Palantir invented it; Amazon Web Services committed USD 1 billion, and OpenAI and Anthropic are building their own teams. Djtal, the operational AI specialist for Swiss companies, applies it to businesses of 20 to 250 employees.
The problem
The AI model works. The problem is integration.
According to MIT (2025), 95% of enterprise AI pilot projects deliver no measurable return (see why AI projects fail). The models are mature and available to everyone. The missing piece is the work of fitting them into the organisation, its data and its real processes. A consultant's report stops short of that work, and so does a vendor demo. Someone has to get inside the system.
Three ways to deliver AI
An embedded engineer writes code inside your systems.
The vendor demo
An impressive pilot on anonymised data that shines in the meeting room and never gets beyond a mock-up.
The embedded engineer (our model)
An engineer who writes code inside your systems and answers for the operational result. The engineer stays until the solution delivers value for you, then hands over the controls.
Our place in this market
Amazon put USD 1 billion into this model. We run it from Fribourg.
Amazon Web Services has committed USD 1 billion to building a team of embedded engineers. OpenAI and Anthropic are building their own. All three reached the same conclusion. The bottleneck in AI is now the engineer who can connect it to the real processes of a company.
These giants work with large global accounts. Swiss companies of 20 to 250 employees are where we work: on the ERP you already run, in French, with management in French-speaking Switzerland and an engineering hub that works every day in our own workflows.
How an engagement unfolds
From a working pilot to a full handover.
01
A pilot running within the first week
We pick one of your processes and connect an initial agent to your tools, so you see something working within a few days, on your own data.
02
We embed in your systems
For eight to twelve weeks we work inside your environment: your ERP, your CRM, your rules, your governance. We write the agent's scope, map your exceptions and build in human oversight wherever the risk calls for it. This is the work that failed pilots skipped.
03
We hand autonomy over to you
At the end you have a working agent, a written operating guide, a trained team and oversight that you run without us.
Customer Zero
Nine AI agents already run Djtal's own processes.
We call this Customer Zero: being our own first customer. Each of our nine agents works to a written scope under a human supervisor, and our AI-native company page explains how that runs day to day. Two of these agents have their own detailed page: the finance agent and the communications agent.
This page was written by Sofia, reviewed by Iris and published after human approval.
Frequently asked questions
What is forward-deployed engineering, and what does a forward-deployed engineer do?
Forward-deployed engineering is a delivery model invented by Palantir. A forward-deployed engineer embeds in the client's real environment, writes code on its systems and data, and stays accountable for the operational result until the solution is in production. By 2026 the model has become the linchpin of enterprise AI. Amazon Web Services committed USD 1 billion to it, and OpenAI and Anthropic are building their own teams of embedded engineers. Djtal scales the model to fit Swiss companies.
How is forward-deployed engineering different from a conventional IT project?
A conventional project separates the person who advises, the person who codes and the person who runs the system, and the integration into your organisation falls through the gaps between them. That is the number one cause of failure. According to MIT (2025), 95% of AI pilot projects deliver no measurable return, for lack of integration into real workflows. Djtal's embedded engineer closes those gaps. One person is accountable for the result, inside your systems, from day one until your team can run it alone.
Do you need to be a large company to benefit from forward-deployed engineering?
No. Djtal applies the model to Swiss companies of 20 to 250 employees. Amazon, OpenAI and Anthropic send their engineers to large global accounts, well out of reach of companies that size. Djtal works close by, in French, on your existing ERP or CRM, whichever it is.
How does Djtal avoid making clients dependent on it?
By treating autonomy as a deliverable of the engagement. At the end of a Djtal engagement, the agent, its documentation and its oversight are in your hands. We reuse our own tooling from one engagement to the next, which shortens delivery times and spares you being billed for bespoke work every time. When you call us again, it will be because you want to.
Which systems do you work on?
Djtal works on the systems you already have. Zoho is its long-standing area of expertise (Authorized Zoho Partner since 2017), but a Djtal agent also connects to Odoo, Microsoft Dynamics or any other system, or works independently of any platform, directly on your emails, documents and spreadsheets. We already work hands-on in the tools where the agent has to act, whereas a purely AI-focused provider is still learning the ropes.
How does an engagement start?
Djtal starts with a scoped audit that identifies the processes that will pay off and weeds out the ideas that only look good, then builds a pilot designed for production. You decide whether to extend the engagement on the strength of a measured result, never on a promise.
Have a process in mind? We start by proving it works.
We scope the process, then an engineer embeds in your systems and has a pilot running on your data within the first week, designed for production. The decision to extend is yours, and you make it with the figures in front of you.
Last updated: