Forrester's Five-Category AI Investment Framework Doesn't Bet on How Autonomous Agents Get

Forrester analysts Charles Betz and Manuel Geitz built a framework around a specific bet: five AI investment categories that pay off “regardless of how autonomous AI technology evolves.” Their argument is that most enterprises are structuring AI budgets around a guess about where agentic capability lands in two years — and that guess is the wrong basis for spending decisions. The five categories: Identity (agent inventory systems with distinct identities, known sponsors, bounded permissions, and explicit lifecycles — application portfolio management, applied to agents); Capability (building agents on existing platform engineering and reusable business services instead of from scratch, illustrated with an example from Lendi); Meaning (semantic alignment through metadata, ontologies, and knowledge graphs, to avoid what the authors call “drug interaction” effects between conflicting AI systems); Confidence (continuous evaluation via guardrails and LLM-as-judge techniques, replacing after-the-fact audits); and Control (feedback loops giving real-time visibility into agent cost, behavior, and risk, with machine-speed intervention capability).

The through-line is bounded autonomy with governance, not maximum autonomy — the same five investments hold up whether the next two years bring fully autonomous agents or merely more capable software that still needs a human in the loop.

That governance-first posture shows up in the numbers, too. Highspot’s CIO tech-stack guide, published a day earlier, found only 53% of enterprise revenue leaders report consistent execution outcomes despite heavy AI investment, and 42% of GTM leaders cite fragmented systems as their top barrier — while organizations with successful AI initiatives invest four times more in data and governance foundations than their peers. Forrester’s framework and Highspot’s data are pointing at the same gap from different angles: the enterprises actually seeing returns aren’t the ones with the most autonomous agents. They’re the ones who built the governance layer first.