AI Opportunity and Risk Diagnostic
Review the workflows, value, data access, failure cost and governance needs before money is committed to a build.
Outputsworkflow map/opportunity and risk register/prioritised 90-day roadmapAI systems and transformation
AI systems should survive contact with the business.
I help leadership teams decide what to change, build governed workflows and create the operating rules needed to use AI responsibly.
Strategy and adoption
Governed workflows and evaluation
Up to three days a week
01
Strategy and adoption
Turn disconnected experiments into a prioritised operating roadmap.
02
Agentic workflows
Connect models, tools and business processes around a clearly owned outcome.
03
Evaluation and governance
Define what good looks like and stop unsupported outputs becoming business facts.
04
Private and local AI
Choose cloud or local processing around the information and risk involved.
In short
What does an AI systems and transformation consultant do?
Thomas Howcroft is an independent AI systems and transformation consultant who helps leadership teams decide what to change, build governed workflows and set the operating rules needed to use AI responsibly. The practice covers four capabilities: strategy and adoption, agentic workflows, evaluation and governance, and private or local AI. Engagements take five shapes: an AI Opportunity and Risk Diagnostic of about two weeks, a Governance and Evaluation Sprint of two to three weeks, a Governed Agentic Pilot of four to eight weeks, a Fractional AI Systems Lead arrangement of one to three days a week, and scoped Leadership and Team AI Training. The method runs through diagnosis, architecture, build and evaluation, then handover and adoption, and the current standard requires human approval before anything is sent or published. The work suits founder-led and mid-market B2B, technology, professional-service and creative organisations. Creative production stays with XO3D, and SEO or AI-search delivery runs through TomHow.
Ways to work together
What engagements does Thomas Howcroft offer?
Each engagement has a defined output, boundary and handover.
Governance and Evaluation Sprint
Review an existing pilot or live workflow, define acceptance criteria and test the failure modes that matter to the business.
Outputsevaluation suite/failure map/operating controls and recommendationGoverned Agentic Pilot
Take one valuable workflow from scope through implementation, evaluation and a usable handover. Use the 16-question AI Pilot Checklist to test whether the pilot is ready to become a dependable system.
Outputsworking workflow/acceptance tests/documentation and handoverFractional AI Systems Lead
Own the roadmap, important technical decisions, governance, delivery oversight and team adoption without a full-time hire.
Scopedefined decision ownership/weekly operating cadence/monthly evidence reviewLeadership and Team AI Training
Practical training built around the organisation's real work, risks and operating rules rather than generic tool demonstrations.
Outputstailored sessions/working rules/practical adoption planSelected evidence
What AI systems has Thomas Howcroft actually built?
The complete portfolio includes internal systems, live applications, public websites and paper-only research. Each case states the evidence and the limit.
A governed AI operating system across working businesses.
Connected tools support research, reporting, publishing, approvals and internal dashboards across active operations.
Read the full caseEvidenceworking interfaces/evaluation and incident loop/human approval for external actionsA private founder dashboard built around decisions.
Today, Briefing, Work, Brand, Health, Review and mobile approvals reduce noise without exposing personal records.
Read the full caseEvidencedesktop and mobile surfaces/honest stale states/exact approval packetsRole-based operational visibility for a working studio.
An internal platform connects project truth, studio schedules, reviews and management decisions while restricting each view by role.
Read the full caseEvidencecontrolled project source/role-shaped views/automated privacy checksA private AI operating system across desktop, Android and wearable devices.
Role-shaped surfaces combine approvals, voice, native notifications, health sources, cross-device state and designed recovery.
Read the full caseEvidencesigned build/physical-device checks/recovery and deep-link proofOvershove connects a working application to governed growth.
The live poker beta, public website, search estate, creative system and AI content operations share explicit quality and approval gates.
Read the full caseEvidencelive beta/offline-aware logging/search and publishing controlsMethod
What is the delivery method?
The work moves from an owned business problem to a system the team can understand and operate.
Diagnosis
Define the workflow, decision, baseline, sponsor, evidence and failure cost.
Architecture
Set the data boundaries, tools, ownership, approval points and acceptance criteria.
Build and evaluate
Implement the smallest useful workflow and test it against known failures.
Handover and adoption
Document the operating rules, train the people involved and agree how evidence is reviewed.
Fit
Is this the right fit for my business?
I work with founder-led and mid-market B2B, technology, professional-service and creative organisations.
Good fit: a defined workflow or decision, a named sponsor, access to the relevant people and data, and a willingness to evaluate before scaling.
Not a fit: full-time employment, generic staff augmentation, pure model research, specialist ML engineering, or external automation without human accountability.
Creative production work belongs to XO3D. Technical SEO and AI-search visibility work runs through TomHow.
Questions
What do buyers usually need to know?
Scope, responsibility and limits are agreed before work begins.
Do you advise or build?
Both. An engagement can stop at strategy or continue into workflow design, implementation, evaluation and handover. The scope states exactly what I own. In practice that means an AI Opportunity and Risk Diagnostic of about two weeks, a Governance and Evaluation Sprint of two to three weeks, or a Governed Agentic Pilot of four to eight weeks, each with a defined output, boundary and handover.
Do you work with internal technical teams?
Yes. I can lead a defined workstream, work alongside an internal team or provide independent direction and evaluation. The method runs from diagnosis and architecture through build and evaluation to handover and adoption, so documentation, operating rules, training and an agreed evidence review are part of the work rather than an afterthought.
Do you automate external actions?
My current standard is human approval before anything is sent or published. Drafting and preparation can be automated, and approval points are set during architecture alongside data boundaries, tools, ownership and acceptance criteria. External automation without human accountability is not a fit for this practice.
How do you handle confidential information?
Data handling and system boundaries are agreed before work begins. Cloud services are used where appropriate, with local processing available where information needs to remain private. Choosing cloud or local processing is treated as a design decision about the information and the risk involved rather than a default setting.
Are you available full time?
No. I take remote UK projects and part-time contracts for up to three days a week. Fractional AI Systems Lead work runs at one to three days a week with defined decision ownership, a weekly operating cadence and a monthly evidence review. Full-time employment and generic staff augmentation are not a fit.
Can you guarantee ROI, rankings or AI citations?
No. I define measurable objectives, report the available evidence and state the limits. Acceptance criteria are agreed during architecture and tested against the failure modes that matter to the business, so progress is judged against evidence rather than a promised percentage. Each case study states its evidence and its limit for the same reason.
Discuss a project
How do I start an AI systems project?
I review every brief myself. If the fit is right, I will suggest a short introductory call. If it is not, I will say so.
Send the current workflow, what needs to change, the sponsor, the timing and the likely investment range.
Remote, UK
Projects and fractional work
Up to three days a week