Roadmaps and scaling
Scaling is where AI value is won or lost. This module turns your priorities into a roadmap and a scorecard leadership can steer by.
Learning objectives
- Build a three-horizon AI roadmap
- Plan what pilots need to reach production
- Manage AI as a portfolio
- Design a leadership scorecard
6.1 A roadmap in three horizons
| Horizon | Focus | Typical outputs |
|---|---|---|
| Now (0–6 months) | Foundations and quick wins | Approved tools, policy, literacy training, 2–3 pilots, data audit |
| Next (6–18 months) | Scale what works; redesign key processes | Production systems, shared platform, metrics dashboard |
| Later (18+ months) | Strategic bets and new offerings | AI-enabled products, deeper automation, agents in core processes |
6.2 From pilot to production
Many organisations get stuck after the pilot. Scaling needs things pilots skip: integration with core systems, security review, support and on-call, monitoring, training for all users, updated processes and job descriptions, and a budget line that doesn't end with the pilot. Plan these before the pilot finishes, so a "go" decision can move quickly.
6.3 Manage a portfolio, not projects
- Balance quick wins, process transformations and strategic bets
- Review the whole portfolio quarterly: scale, iterate or stop each item
- Stop things — killing weak pilots frees money and people for strong ones
- Reuse components (platforms, prompts, evaluation sets) across use cases
6.4 Reporting value
A simple quarterly AI scorecard for leadership:
| Value | Outcomes delivered vs business case (cost saved, revenue, CSAT), by use case |
| Adoption | Active users, usage trend, coverage of target processes |
| Portfolio | Use cases by stage; decisions made (scaled, stopped) |
| Risk | Incidents, open risk actions, compliance status |
| Capability | People trained, skills gaps, platform maturity |
6.5 Keep the strategy current
AI capabilities, prices and regulations change every few months. Review the strategy at least twice a year: what new capabilities open up use cases that weren't feasible? Which assumptions in our business cases have changed? What did we learn from what we stopped? A strategy that doesn't change is probably not being used.
Bring the course together on one page: three priority outcomes; your top use cases by horizon (now / next / later); build-buy decisions; operating model; first governance controls; and the five scorecard measures you'll report quarterly.
See a worked example
Example headings: Outcomes — cut cost-to-serve 15%, grow repeat purchases 10%, zero AI-related compliance breaches. Now — support drafting pilot, approved assistant for all staff, AI literacy for managers. Next — scale support drafting, personalised bundles. Later — conversational shopping assistant. Operating model — hub and three spokes. Scorecard — cost per contact, repeat rate, active users, incidents, staff trained.
Key takeaways
- Now, next, later: foundations, scale, strategic bets.
- Plan production needs before the pilot ends.
- Review the portfolio quarterly — and stop things.
- Report value, adoption, portfolio, risk and capability.
- Revisit the strategy at least twice a year.