New·Argos now detects usage deviation across 100+ model endpoints See how →
Home / Platform / AIM framework
Method

AIM: how an engagement runs, phase by phase.

Each phase ends where the next one can start: an assessment you can act on, a use case running in production, then the routine handled and the exceptions escalated. Most clients run the full cycle in six to nine months, and you can start at whichever phase fits.

Phase 01 · 3 weeks

Assess

Where are we, honestly?

  • Scanning agents deploy read only in your estate
  • Evidence based maturity score per business unit
  • Full inventory: owner, KPI, run cost, status
  • Use cases ranked on value, readiness, exposure
  • Argos spend baseline, signed by Finance

Gate: a ranked backlog where every use case has a sponsor, a baseline and a target.

Phase 02 · 12 weeks to first outcome

Implement

Can we ship one outcome and prove the pattern?

  • Spec signed before a line of code
  • Voyager agents build; engineers review every merge
  • Shadow → approve → bounded autonomy on evidence
  • Reviewers trained and in the loop from week one
  • Pattern captured for the next business unit

Gate: an agent in approve mode with a measured business outcome.

Phase 03 · ongoing

Mature

Can it run better with fewer hands on it?

  • Tiered autonomous maintenance: L1 resolve, L2 triage, L3 forensics
  • Drift, quality and fairness monitored continuously
  • Argos routing and deviation detection portfolio-wide
  • Board ready governance pack from live evidence
  • Pattern library so the next build starts ahead

Graduation: your team runs it. We move from running to advising.

AI Maturity Model

Most enterprises think they're Level 2.

The evidence usually says otherwise. Assess scores each business unit against the model, and the score decides which phase starts first and what a pod looks like.

Level 1 · Foundational

“We are experimenting with AI.”

Opportunistic proofs of concept, siloed data, limited KPI linkage, high cost and low realised value. Most use cases never leave PoC, and nobody can say what AI costs to run.

Level 2 · Scaled

“We are deploying AI.”

Productionised models and MLOps, clear KPI alignment, value trapped in silos. Moderate reuse. Costs known per project, never per outcome.

Level 3 · Autonomous

“AI runs our business.”

Self-learning loops, agentic execution, tied to revenue and risk at CXO level. High decision automation and a cost per outcome that trends down every quarter.

Side by side

What changes between phases.

Engagements can start at any phase. Most clients run the full cycle over six to nine months.

AssessImplementMature
Question answeredWhere are we, honestly?Can we ship one outcome?Can it run with fewer hands?
Duration3 weeks12 weeks to first outcomeOngoing, for as long as it earns its place
Who is in the podAssessment lead, domain SME, data lead + scanning agentsArchitect, engineers, operators + Voyager build and eval agentsSRE, change lead, FinOps analyst + Argos and maintenance agents
Gate to next phaseRanked backlog with owners and baselinesApprove mode agent with measured outcomeOutcome, adoption and cost per outcome on the console
Agents involved6 Assess agents22 Implement agents9 Mature agents
How we are paidFixed fee, credited against ImplementBase plus share of the outcomeShare of outcome and cost reduction

You do not have to run all three. Most clients start where their gap is and complete the full cycle in six to nine months. If the prioritisation work is already done, we start at Implement.

Let's talk

Tell us the number you need to move.

A 45-minute working session with an operator who has run the kind of work you are describing. You will get an honest read on where your programme stands and what it would take to move it.