OrqiveAI Take the assessment
Services Process mining How it works Agentic AI vs Copilot Glossary FAQ Take the assessment
How it works · 5 phases

How an agent gets built, from assessment to handover

Short answer

Five phases, about 90 days, ending with a governed agent running on live volume and owned by your team. Readiness assessment (1–2 weeks), agent blueprint (2 weeks), environment preparation (2–4 weeks), build and pilot (4–6 weeks), then handover and measurement.

Agents are built inside your own tenant. Nothing depends on Orqive infrastructure, and every agent, template, and integration pattern is yours at the end.

  1. Phase 01 · 1–2 weeks

    Readiness assessment

    We grade your environment L1–L4 across six dimensions with the OARI, map your core processes against the APQC Process Classification Framework at L3–L5, and produce a gap register that names where agents deliver the fastest return. This is where a lot of assumed readiness turns out to be a data or ownership problem instead.

    Where the event data supports it, process mining runs alongside the walkthroughs so the gap register is priced from your own logs rather than estimated in a workshop.

    You give: process walkthroughs Optional: process mining You get: OARI score + gap register
    Tier 1
  2. Phase 02 · 2 weeks

    Agent blueprint

    For your top three to five gaps we specify each agent: what triggers it, which systems it touches, the decision and exception rules it applies, where it escalates to a human, what it must log for audit, and the metric that says it worked. The blueprint is build-ready — your team, your Microsoft partner, or ours can execute against it.

    You give: exception rules, approval hierarchy You get: agent design pack
    Tier 1
  3. Phase 03 · 2–4 weeks

    Environment preparation

    The blueprint names its own prerequisites, and this phase closes them: master-data cleanup with named owners, the integration and Dataverse connections the agent will call, security roles, and the governance model for logging and human-in-the-loop review. Where prerequisites are already met, this phase runs in parallel with the build.

    You give: tenant access, data owners You get: agent-ready environment + governance model
    Tier 1/2
  4. Phase 04 · 4–6 weeks

    Build and pilot

    We build the first agent in Copilot Studio, Power Platform, and Azure AI inside your tenant, then run it alongside the manual process on live volume. The pilot exists to tune exception thresholds against real cases — the ones your documented rules did not anticipate — before anything is cut over.

    You give: a process owner for pilot review You get: working agent on live volume
    Tier 2
  5. Phase 05 · ongoing

    Handover and measurement

    Agents, templates, prompt libraries, integration patterns, and process documents transfer to you. A named owner is trained to run and change them. Value reporting goes up against the metrics agreed in phase two — hours recovered, cycle time, exception rate — and the next workflows are prioritised from the gap register.

    You give: a named internal owner You get: full IP transfer + value reporting
    Tier 2
What the phases assume

Three things carry the timeline

When a 90-day engagement slips, it is almost never the build. It is one of these three, discovered late.

01

Documented process

An agent runs steps. Person-dependent routines have to be standardised before they can be executed or audited.

02

Written exception rules

Thresholds and owners held in someone's head cannot be automated. Writing them down is phase-two work.

03

Owned master data

Duplicate customers and stale items become automated errors at scale. Cleanup precedes deployment.

FAQ

What clients ask before phase one

On timeline, effort, tenancy, and what you keep.

How long does the first agent take to go live?

About 90 days from the readiness assessment to a governed agent running on live volume. Assessment and blueprint take two to four weeks, environment preparation two to four weeks, and build and pilot four to six weeks, with phases overlapping where prerequisites allow.

How much of our team's time does an engagement need?

Expect a process owner for about four hours a week during assessment and blueprint, and an IT or ERP contact for tenant access, security roles, and integration questions during build. Orqive does the design and build work; you supply process knowledge and access.

Do you build inside our tenant or yours?

Yours. Agents are built in your own Azure, Copilot Studio, Power Platform, and Dataverse environment so nothing depends on Orqive infrastructure and everything remains under your security, licensing, and audit controls.

What do we own at the end?

Every agent, template, prompt library, integration pattern, and process document produced during the engagement. A named owner is trained to run and modify them, and no part of the agent layer requires Orqive to keep operating.

Next
Phase oneTake the OARI assessment18 questions, about 10 minutes, graded L1–L4 across six dimensions. ComparisonAgentic AI vs CopilotWhere Copilot stops and an agent starts, side by side. GlossaryAgentic AI, definedPlain definitions for the terms in every agent conversation.