AI Systems
Decide where AI actually fits
Pressure to adopt, with no way to prioritize
Leadership is being asked what the company is doing about AI. The result is usually a scatter of pilots: a chatbot nobody uses, a licence spend nobody can account for, and a few teams quietly pasting internal data into consumer tools. Nothing connects to a workflow that matters. The missing piece is an honest assessment of which processes are genuine candidates, which are blocked by data or accountability problems, and what order to build in.
What we build
A sequenced plan for where AI belongs in your business, and where it does not.
Workflow assessment
We look at the work, not the technology, and score candidates on evidence.
- Process mapping with the people who do the work
- Volume, repetition and exception profile per workflow
- Data availability and quality for each candidate
- Accountability review: who owns the decision today
- A ranked shortlist with the rejected candidates and the reasons
Readiness and gap analysis
Most AI projects fail on plumbing and ownership rather than on models.
- Systems-of-record review and integration feasibility
- Data quality, structure and access gaps
- Identity, permissions and retention constraints
- Internal skills and maintenance capacity
- Shadow AI usage already happening in the business
Governance and usage policy
Rules that are specific enough to follow, written for your actual tooling.
- Approved and prohibited use cases, in plain language
- Data-handling rules for internal and customer information
- Human-review requirements by decision type
- Disclosure standards for customer-facing systems
- Review cadence as tools and regulation change
Sequenced build plan
A first project chosen to prove or disprove the approach quickly.
- A narrow first build with a defined success test
- Dependency order across later phases
- Ownership, maintenance and cost expectations per phase
- Decision points where a phase should be cancelled rather than continued
How the work runs
Discovery
Interview the teams doing the work, inventory systems and data, and surface the AI usage already happening unofficially.
Assessment
Score candidate workflows on repetition, data readiness, risk and accountability, and write down what we are ruling out.
Plan
Produce the governance rules, the sequenced roadmap, the ownership model and a realistic cost and maintenance picture.
First build
Deliver the narrow first project against its success test, then review whether the rest of the sequence still makes sense.
What you end up with
- A ranked shortlist with explicit reasons for what was rejected
- Governance rules specific enough for staff to follow
- A first build narrow enough to prove or disprove the approach
- Leadership can explain the plan without relying on vendor language
Typically in scope
- Core systems of record
- Data warehouse and reporting layer
- Identity and permissions
- Existing automation and integration tooling
- Collaboration and documentation platforms
- Security and retention policy
Where this comes up most
What we will not automate
Scope boundaries agreed before the build starts, not discovered after launch.
Headcount decisions
We assess workflows, not people. Any restructuring or role change is a leadership decision, and we will not present a staffing cut as a technical finding.
Accountable decisions
Workflows where a named person must answer for the outcome stay human-led. We will recommend against automating them even when it is technically possible.
Buying before scoping
We do not recommend a platform or licence before the workflow is defined, and we hold no vendor partnerships or resale arrangements that would make that recommendation less than independent.
Where we have built this
Each one is written up in full, with its delivery stage stated plainly. Clients are described by industry only.
AI Transformation, answered
What teams usually ask before scoping this.
The deliverable is a ranked workflow shortlist, a governance policy and a sequenced build plan with ownership and cost attached. It is written to be executed, and we would rather hand it to your own team than have it shelved.
Then we say that, and point at the data, integration or ownership gap that needs closing first. That is a cheaper finding than a failed build.
Only after the workflow is defined, and only as a category plus a shortlist with trade-offs. We hold no partnership or resale arrangement with any vendor.
We find them during discovery and treat them as evidence of real demand. Banning them without offering a safe alternative tends to push the usage further out of sight.
You do. The documentation, the assessment detail and the reasoning behind each decision are yours, including the parts that argue against building.
Related services
Scope a ai transformation build
Book a strategy call. We'll map the workflow, find the real bottleneck, and tell you honestly what should and should not be built.