The AI Service Lifecycle, on Halo

When six useful agents become one management problem.

Each agent was a good decision on its own. None of them is the problem. The portfolio is. The lifecycle is how you plan, prepare, introduce, operate, assure and evolve AI services so they run as part of your service portfolio.

Take the scorecard See the assessment

Fixed-price assessment: £4,950 + VAT, two weeks, agreed before we start.

AI service recordAGT-005 · illustrative

Password reset agent

Owner: Security Lead
Authority: Supervised
Status: Live

Purpose

Resets staff passwords after an identity check, in the flow of the service desk, without an analyst.

Control envelope

May

Verify through MFA
Use the approved runbook

May not

Change groups
Touch privileged accounts

Sees

The requester’s directory record only

Written to the service record

09:41Reset completed. MFA confirmed.
09:38Privileged account escalated to analyst.
TueContext change CHG-2231 released.

Last assurance review: 12 days ago Next: 18 days

Then three questions arrive.

From people who never asked for an agent, and who cannot be answered with a demo.

Finance asks

What are these agents costing us, and what are they returning?

Spend is spread across platforms and projects, and nobody owns the total.

Security asks

What data can they see, and what are they allowed to do with it?

Authority was set during configuration, not stated as a control.

Operations asks

What happens when one gets it wrong at 2am?

There is no incident route for agent behaviour, so it goes unreported.

Building the agent is the easy part.

A project ends at handover. A service has to keep working safely, at a cost you can explain.

A project

Build the agent

Scoped, configured, demonstrated and handed over.

Use case and prompt
Configuration and testing
Demo and handover

Success means it works.

A service

Run the AI service

Owned, catalogued, governed, measured and reviewed.

Named owner and defined authority
Introduction, change and incident routes
Assurance, review and retirement

Success means it keeps working safely, at a cost you can explain.

Six stages, from decision to retirement.

One agent, threaded through all six. Choose a stage to see what changes on its record.

What this stage decides

Which AI services are worth running. Use case, business owner, value measure and risk position are defined before anything is built.

Deliverable

Opportunity case

Halo artefact

AI Service Register entry

Password reset agentAGT-005

Status: Proposed
Owner: Security Lead
Authority: Not yet set

Register entry
Control envelope
Introduction record
Catalogue entry
Under change control
Incident route open
On the assurance dashboard

Password resets are the highest-volume request on the desk. Value is measured by resets resolved without an analyst. Identity risk requires an approval gate.

Illustrative. The stages and artefacts are real; the agent is an example.

What this looks like inside Halo.

Three layers sit under every AI service. The lifecycle is built as working artefacts in the platform you already run.

Layer 03

Live agent

Doing the work in the flow of the service, with every action recorded.

Layer 02

AI control envelope

The context, abilities, tools and authority defining what the agent may do.

Layer 01

Service context

The Halo Service Automation Framework: Service Data Model and Target Operating Model.

Six artefacts, built in your Halo environment.

AI Service Register

Every agent, its owner, purpose and authority in one record.

Serves Plan and Operate

AI Introduction Workflow

The route a new agent takes from proposal to live service.

Serves Introduce

Agent Change Control

Prompt, context, tool and authority changes handled as change.

Serves Operate

AI Incident Workflow

A reporting and resolution route for incorrect agent behaviour.

Serves Assure

AI Service Catalogue

What the AI service does, for whom and what to expect.

Serves Operate

Assurance Dashboard

Quality, containment, escalation and authority, reviewed regularly.

Serves Assure and Evolve

The AI Service Lifecycle brochure is the long version of this page.

Four levels. Most organisations are at one or two.

Level 1

Experimenting

What exists?

Level 2

Operationalising

Can we trust it?

Level 3

Governed

Can we scale it?

Level 4

Scaled

Is it creating value?

Score yourself in five minutes

AI Service Readiness and Portfolio Assessment.

A fixed-price, two-week engagement that tells you what AI is actually running, how mature it is, and what to do in the next ninety days.

The scorecard is self-scored and tells you where you think you are. The assessment is consultant-scored against your live estate and tells you what is actually running, what it costs, and what to do in the next ninety days.

£4,950 + VAT

Fixed, agreed before we start.

Delivery: Two weeks from kickoff to findings.
Your input: Four to six hours, including a 90-minute kickoff and two working sessions.
Scope: Up to ten AI services in one Halo environment.

What we look at

Every agent, feature and pilot live or in flight
Service data, knowledge and context
Authority, approvals and audit position
How AI enters, changes and leaves service
Where the platform can carry the lifecycle without new tooling

Book a briefing Take the scorecard first

What you get

AI Service Register: a single view of every live and proposed AI service
Maturity level: a scored position across the six lifecycle stages
Risk and value classification: a clear view of exposure and expected return
Readiness gap report: the controls and capabilities missing today
Prioritised 90-day roadmap: sequenced actions for the next ninety days

Three design-partner places

£2,950 + VAT in exchange for an agreed reference and case study after delivery.

The assessment is the entry point, not the whole road.

Now

Assessment

Then

AI Service Foundation

Build the six lifecycle artefacts in your platform.

Ongoing

AI Service Assurance

Run quality, authority, incident and service reviews.

As you grow

AI Service Evolution

Expand, improve, restrict or retire on evidence.

Questions we get asked about the lifecycle.

What counts as an AI service?

Any agent, AI feature or pilot that answers or acts in the flow of a service. If it can affect a ticket, it is in scope.

Does this need new tooling?

No. The lifecycle is built as artefacts in your existing Halo environment, under the Halo Service Automation Framework and its Service Data Model.

Is this a new framework we have to adopt?

It applies familiar introduction, change, incident, catalogue and review disciplines to agents. Each agent gains a record, owner and authority state.

What if we have more than ten AI services?

Launch scope covers up to ten services in one Halo environment. Larger estates are scoped before the price is agreed.

How is the scorecard different?

Who does the work?

The team behind HaloNinjas, who run service desks and Halo estates day to day. The people in the briefing are the people on the engagement.

Start with where you actually are.

Five minutes on the scorecard tells you your level and your three biggest gaps.

Take the scorecard

Next step

Book a 30-minute briefing.

A briefing, not a pitch.

Book a briefing AI Service Lifecycle brochure