Value, feasibility, risk and adoption, in the same conversation from day one.
How we work
A stage-gated, co-delivery model. At each gate, client and consulting leads explicitly decide to continue, revise, pause or stop. This prevents sunk-cost momentum from replacing evidence.
Six commitments that shape every engagement
Delivery principles
Value before volume
Prioritize a small number of measurable opportunities.
Evidence before scale
Test value, quality, safety and behavior, not technical feasibility alone.
Humans stay accountable
Design oversight and escalation around consequence and uncertainty.
Reuse before rebuild
Work with the client's platforms and standards where they are fit.
Open by design
Use clear architecture, portable knowledge and explicit vendor trade-offs.
Transfer from day one
Pair, document and teach throughout, not only at handover.
Eight stages, three decision gates
The journey
STAGE 0
Connect & protect
Establish fit and a safe basis for discussion.
Outputs: Mutual NDA where needed, initial hypothesis, stakeholder map and next-step proposal.
STAGE 1
Onboard & align
Agree outcomes, scope, ways of working and access.
Outputs: Charter, success measures, governance cadence, access/data plan and RAID log.
STAGE 2
Discover & assess
Understand work, users, data, platforms, controls and constraints.
Outputs: Current-state map, readiness baseline, opportunity backlog and risk view.
DISCOVERY GATE: CONTINUE, REVISE, PAUSE OR STOP
STAGE 3
Prioritize & design
Choose the highest-value viable intervention and define its operating context.
Outputs: Value case, experience concept, architecture, control plan and delivery backlog.
STAGE 4
Prove value
Test business value, technical feasibility, safety and user behavior.
Outputs: Pilot, evaluation results, learning log and scale/no-scale decision.
VALUE-PROOF GATE: SCALE, REVISE, PAUSE OR STOP
STAGE 5
Build & integrate
Engineer the production capability and connect it to real systems and controls.
Outputs: Tested product, integrations, telemetry, documentation and release evidence.
PRODUCTION-READINESS GATE: RELEASE, REVISE, PAUSE OR STOP
STAGE 6
Adopt & transfer
Embed the capability in roles, decisions and daily work.
Outputs: Training, champions, communications, support model and ownership handover.
STAGE 7
Operate & improve
Sustain quality, manage risk and compound value.
Outputs: Monitoring, service reviews, improvement backlog, quarterly value report and exit plan.
Three connected levels keep decisions close to evidence
Governance & co-delivery
| Level | Participants | Cadence | Focus |
|---|---|---|---|
| Executive steering | Executive sponsor, accountable business owner, CIO/CDO/risk as relevant, consulting partner | Monthly or at major gates | Value, risk, investment, dependencies and decisions |
| Product & value team | Product owner, process owner, data/technology leads, change lead, consulting lead | Weekly | Priorities, measures, user feedback, readiness and blockers |
| Delivery squad | Client SMEs and cross-functional consultants: product, process, data, engineering, UX, risk/change | Daily collaboration; 1–2 week sprints | Design, build, evaluate, integrate, document and transfer |
Our responsibilities
- Challenge the problem and value hypothesis.
- Bring senior business, product, data and engineering capability.
- Make trade-offs and risks visible early.
- Design for security, adoption and operations.
- Document decisions and transfer capability.
- Report evidence honestly, including failed assumptions.
Client responsibilities
- Name an accountable sponsor and empowered product owner.
- Provide timely SME, user, data and system access.
- Engage security, legal, risk, HR and procurement early.
- Own business decisions and organizational change.
- Validate measures, controls and release decisions.
- Plan internal ownership before production.
A credible first step and a visible route to scale
Engagement formats
| Format | Best for | Indicative shape | Core output |
|---|---|---|---|
| Executive AI briefing | Leadership alignment and decision framing | 1–2 facilitated sessions | Shared context, opportunity themes and decisions required |
| AI opportunity & readiness sprint | Moving from broad ambition to priorities | 2–4 weeks | Readiness baseline, prioritized use cases and 90-day action plan |
| Value proof | Testing value, feasibility, risk and adoption before scale | 6–10 weeks | Working pilot, evidence scorecard and scale/no-scale decision |
| Build & scale program | Production delivery across systems and teams | Multi-phase, typically 3–9 months | Production capability, controls, adoption and ownership transfer |
| Embedded AI squad | Adding senior cross-functional capacity to a client program | Dedicated team, sprint cadence | Prioritized delivery backlog and integrated capability |
| Fractional CAIO / advisory | Ongoing executive leadership without a full-time role | Monthly retainer | Portfolio direction, governance, vendor and investment decisions |
| Managed AI improvement | Operating and improving live AI products | Ongoing service with agreed levels | Monitoring, evaluation, optimization, support and value reviews |
Durations are indicative. Actual scope depends on access, data readiness, risk level, integrations, procurement and client availability.
Operate & improve: sold as stewardship, run on evidence
Aftercare
Aftercare is active product stewardship, not a vague support promise.
Launch stabilization
Typical period: first 30 days after release.
- · Priority incident and user support
- · Usage, latency, cost and quality observation
- · Knowledge and retrieval tuning
- · Workflow and escalation refinements
- · Daily/weekly launch review as appropriate
Managed improvement
Typical period: ongoing, with defined service levels.
- · Automated and human quality evaluation
- · Model, prompt, agent and knowledge changes
- · Security and misuse monitoring
- · Adoption and business-value analysis
- · Prioritized improvement releases
Minimum operating scorecard
| Dimension | Illustrative measures |
|---|---|
| Business value | Cycle time, avoided effort, conversion, service level, decision latency or revenue influence |
| Adoption | Eligible vs. active users, repeat use, task completion, abandonment and satisfaction |
| Quality | Task success, groundedness, accuracy, escalation, false positives/negatives and human acceptance |
| Risk & safety | Policy violations, sensitive-data events, harmful outputs, control failures and incidents |
| Reliability | Availability, latency, integration failures and recovery time |
| Economics | Cost per task/outcome, token/model cost, support effort and infrastructure utilization |
Every quarter we reconfirm the business baseline, benefits, risks, user behavior, technical health, cost and ownership, and agree what to improve, scale, retire or redesign next.