AItlas · The left upright
AItlasUnderstand the business before building it.
AItlas is the consulting platform. It runs the engagement, produces the analysis, and records the reasoning behind every recommendation — ending in an application requirement the client can approve.
- Responsible for
- Understanding the client's business and producing the advice
- Not responsible for
- Choosing technology, building, deploying — that is Anvil
- Hands to Data Weaver
- The approved application requirement
01The consulting workflow
Each analysis grounded in the one before it.
Analysis runs in a declared dependency chain, in the background. If it stops for an approval, it resumes from that point rather than restarting.
- 01
Client dossier
Start from the company name and the client's own material. Three ingest passes build the profile and collect evidence, hypotheses and interview questions.
- 02
Company research
A research brief on the client and its market, grounded in real sources rather than model recall.
- 03
Competitive landscape
Competitors, positioning and a quadrant view.
- 04
Strategy map
Strategic moves, each with its reasoning retained.
- 05
AI technology atlas
A diagnosis, then a catalogue of recommended AI capabilities — plus anti-recommendations, the things not to do.
- 06
Go-to-market plan
Phases, a move-by-phase matrix and delivery control.
- 07
Application portfolio
Capabilities clustered into standalone, buildable applications. What fits nowhere is reported as unplaced; what should not be built is refused, with reasons.
- 08
Application requirement
For each application, a requirement written in the client's own business language.
- 09
Application specification
Five dependent passes — entities, API contract, features, narrative, screens — each reading what the previous one wrote.
- →
Consultant approves · hand-off to Data Weaver
The first hard gate. A person decides the application is ready to build.
02How it reasons
It keeps the reasoning, not just the answer.
The consultant's task becomes review, not production: reading what AItlas produced, correcting what is wrong, and deciding what still needs a conversation with the client.
The reasoning is kept
Every recommendation is recorded with the reasoning behind it, so the understanding that used to leave with the team stays with the engagement.
Evidence and hypotheses stay apart
What AItlas inferred but could not evidence is shown as a hypothesis, with the reason — and enters the dossier only when the consultant adopts it.
It asks rather than guesses
Where two companies share a name, AItlas asks which is meant. Returning nothing is correct; returning the wrong company is not.
Unsupported values stay empty
Below a confidence threshold a field is left empty rather than filled. Figures are never invented.
Grounded, or it says so
When web grounding is unavailable, AItlas refuses rather than returning a knowledge-only answer dressed as research.
The consultant's corrections win
Review screens per analysis, with regenerate and fine-tune per section. A fill-gaps run completes only what is missing, so a correction is never overwritten.
03What the client receives
Four formats. The same content in each.
Every format renders from the same deterministic models, and no model is called at export time — so analysis that was produced cannot go missing on its way to the client.
- Web reportThe in-application summary report and slide deck
- Print reportA paginated report, ready to save as PDF
- DOCXThe full written deliverable
- PPTXThe slide deliverable, with speaker content
Enter the platform
Two doors, one structure.
Consultants work in AItlas. Developers work in Anvil. Data Weaver carries the approved requirement from one to the other.
For consultants
AItlas
- Understand the business.
- Build the strategy.
- Define what should be built.
For developers
Anvil
- Turn the approved requirement into software.
- Design. Build. Deploy. Release.