Enterprise
Your company’s AI.
It learns how your company works, and answers only to you.
- Weeks for a Test Lab: one task, your data, your baseline
- 4
- To roll back to the last release, by design
- 60 s
- A dedicated GPU a month, proposed
- $7,700
- Eigna Box Enterprise, planned
- Q1 2027
What we hear
“AI is a leak risk, and custom AI needs a team we do not have.”
Every conversation starts with four answers. Together they shape the first expert.
- The taskThe one that costs the most time or risk.
- The dataWhat it runs on, and where it must stay.
- What you triedAnd why it fell short.
- Success in 90 daysThe number that would prove it.
Frontier and owned
Use the frontier for broad work. Own the model for your core.
Keep renting
- Email and office work
- Open questions and research
- Generic drafts
Own the model
- The one task that defines your business
- Data you will not give away
- A model that learns it, and stays yours
On narrow tasks, tuned open models beat GPT-4 by 10 points on average across 31 tasks (LoRA Land, 2024).
The Formula, for you
Your work makes an open model yours.
- Base model
- An open model, for example Qwen, Llama or Mistral.
- Your work
- Your files, cases and approved answers.
- Your company’s model
- Trained by us, owned by you, with a Model Deed.
The model is yours, with a signed Deed.
The FormulaHow your people use it
Choose an expert. Review the work. Decide what it learns.
- 01Choose an expertYour team’s expert, and its files.
- 02Review the workIt drafts with sources; your people check.
- 03Decide what it learnsOnly approved examples train the next version.
Payment moves to 45 days 1. Notice to end the contract grows to six months 2.
- 1Supplier terms 2027, approved draft
- 2Framework agreement, section 9
What you will ask
Five objections, answered plainly.
- 01
Data
Our data cannot leave.
With Eigna Local it does not: the model runs on your machines or an Eigna Box on site. With Hybrid and Cloud, data stays in the region you choose. Your data trains only your model.
- 02
Team
We have no ML team.
You do not need one. We deliver it, set it up and run it. Your side names the task and its data owner, scores the work and decides each release.
- 03
Quality and risk
What if a new version is worse?
It does not go live. Every candidate is tested on a sealed set it never saw, against the version you use today; a tie keeps the current one. Roll back in 60 seconds.
- 04
Cost
AI costs rise with every use.
Setup once, then a fixed monthly fee per site or per GPU, never per seat. Proposed: Eigna Local $1,100 a site a month, Eigna Cloud $7,700 a dedicated GPU. For light use of the cheapest rented models, renting stays cheaper.
- 05
Board and regulator
Who is accountable for it?
A named owner, a Model Deed and a release record: who owns the model, what changed and why. Designed to support your compliance work, not to replace it.
Where it runs, for you
Racks on your own servers, or a dedicated GPU in your region.
Eigna Box Enterprise puts racks in your data centre, one Figure in every unit. Your own Apple silicon, NVIDIA RTX or GB10 machines work too.
- Eigna Local
- On your servers or an Eigna Box on site. Data never leaves.
- Eigna Cloud
- A dedicated GPU in the region you choose, with your keys. Data stays in the region you choose.
Proof
Published, losses included.
17.8%
Our first training run cut forecast error 17.8% against its base model. It is not yet ahead of a simple last value baseline, and we publish that too.
Our first client: a forecasting model for trading. It shares founders with Eigna, said up front. Then a media agency, for image and video, and a web agency, for code.
How it grows
From one task to every team.
- Weeks 1 to 4Test LabOne task, your data, your baseline.
- After the gateFirst releaseLive for one team.
- When you chooseThe next taskA new expert on the same base.
- ThenEvery teamOne base, many experts, one Deed each.
Timings are indicative until the first projects run.
Start
Start with one task.
Choose the task and its data owner, agree the Test Lab scope, and kick off with your team.