1. Home
  2. Approach
Approach · Planned on paper

Our approach

We plan AI rollouts the way engineers plan a build: survey the site, lay the foundations, test a section, then complete the job. You see the drawings before anything is switched on, and the measurements after.

Approach · The five layers
S-02 SecurityIdentity, policy, protection
S-04 DataPermissions, labels, clean-up
S-05 AutomationFlows and agents
S-03 CopilotThe AI tools people use
S-06 PeopleTraining and adoption

AI projects tend to fail in one of two ways. Either nothing is prepared, and an AI tool surfaces information it shouldn’t. Or so much is prepared that nothing gets used. We avoid both by working through five layers in order, with a clear finish line at each step.

The five-layer AI stack

We draw every rollout as a stack of five layers. The bottom layers carry the weight of the ones above, so we start there.

  1. Security. Who can sign in, from where, and which AI tools are approved.
  2. Data. What AI can see, whether it’s labelled and whether it’s up to date.
  3. Automation. Processes that run themselves, with a person checking what matters.
  4. Copilot. The AI tools people use every day, such as Microsoft 365 Copilot.
  5. People. The training, champions and habits that turn licences into time saved.

Each layer has a status in your report: ready, in progress or to review. You always know where things stand.

Our principles

How we work

Outcome first

We start with the work you want done faster or better, then choose the technology.

Foundations before features

Security and data come before licences. It’s slower for a week and safer for years.

Small, measured pilots

A few people, real tasks and weekly measurement before anything goes wider.

Honest about limits

We say what AI won’t do well, and when a simpler fix would serve you better.

Plain English

Reports you can share with directors, governors and trustees without a glossary.

Yours to keep

Plans, policies, prompts and documentation belong to you, whoever does the next step.

Engagement method

Four steps, each with a finish line

  1. 01Weeks 01–02

    Discover

    Workshops and a technical review. Finish line: a scored readiness report and a roadmap.

  2. 02Weeks 03–04

    Prepare

    Policies, permissions and protection. Finish line: the security and data layers marked ready.

  3. 03Weeks 05–08

    Pilot

    A small group on real tasks. Finish line: time saved measured and a rollout decision.

  4. 04Weeks 09–12

    Adopt

    Rollout, training and automation. Finish line: everyday use and a first quarterly review.

Questions

Common questions

How do you price your work?
Every engagement is bespoke. After a discovery call we send a fixed quote for each stage, based on how many people are involved, how your data is stored and what’s already in place. You decide whether to continue after each stage.
Are Microsoft licences included?
Licences are separate, because many organisations already have them or qualify for education or charity pricing. We’ll tell you exactly which licences you need.
Why not just switch AI on and see what happens?
Because the risks are front-loaded. If files are overshared on day one, AI can surface them on day one. A few weeks of preparation avoids that, and a measured pilot shows whether the cost is worth it.
Is 12 weeks fixed?
No. It’s a typical plan for a Copilot rollout in an organisation of 50 to 250 people. Smaller organisations often move faster; larger or more complex ones take longer. Your plan is sized after discovery.
What happens after the 12 weeks?
Most clients keep a quarterly review: usage, time saved, new features worth trying and the next three actions. Some add automation projects as they find new processes worth improving.
Talk to us

Start with a readiness assessment.

We look at how you work, what your data looks like and where AI would save the most time. You get a scored report and a plan in plain English, whatever you decide next.

Call us · Mon to Fri, 08:30 to 17:30
01684 215 165
Email
hello@spectrum-ai.uk