Four phases to production software
The client team receives a defined deliverable at the end of each of four phases.
3-day free trial on one matter. No setup fee.
Context AtlasAI reads
The attorney asks
Opportunity map and evaluation criteria, with the playbook positions used as test cases
A prioritized opportunity map
In two to three weeks, the team maps where AI creates value and where AI does not.
Process mapping
The team maps processes with the people who do the work.
System inventory
The team inventories data and systems, with access controls.
Known constraints
The team captures compliance, security and procurement constraints.
Target architecture and roadmap
In two to four weeks, the team delivers milestones, a target architecture, evaluation criteria and an integration plan.
Model strategy
The team sets the solution architecture and model strategy.
Build or buy
The team states each build-or-buy decision with cost and lock-in trade-offs.
Risk register
The team logs data, compliance, model and adoption risks in a register.
Production system per build
In six to twelve weeks per build, the team moves from prototype to production with weekly demonstrations.
Weekly demonstrations
The team demonstrates progress weekly against the evaluation framework.
Security review
The team completes security review, threat modeling and audit logging.
Client-held code
The client receives the code, infrastructure as code, evaluation suite and runbook.
Measured improvement after launch
After launch, the team monitors performance, runs regression tests on production traffic and builds governance.
Regression tests
The team runs regression tests on production traffic.
Model updates
The team updates models and prompts as providers change.
Operation or handoff
The client selects managed operation by AtlasAI or a full handoff.
Questions attorneys ask
Who owns the code and the evaluation suite?
The client owns the source code, the infrastructure as code, the evaluation suite and the runbook. The system runs without AtlasAI and without a lock to one model or one cloud.
Does AtlasAI train models on client data?
No, unless the client explicitly chooses training. The team deploys in the environment of the client and keeps the data of each client separate.
How does the team measure success?
The team defines the evaluation criteria in Phase 2 and delivers the evaluation suite with the system.
Does the team decline projects?
Yes. The team states when a problem is a poor fit, when a simpler solution is better or when a phase requires more time.