// Diagnostic

You are getting squeezed from both sides.

The board wants results. Delivery does not pause.

Leadership is pushing to move faster with AI. Meanwhile you have existing commitments, delivery pressure, and a full plate that does not stop while everyone figures out what "doing something with AI" actually means. You are on the hook for results your organization is not set up to deliver. There is no shared way to decide where to focus, how to govern it, or how to keep any of it alive once the pressure hits.

This diagnostic places you on the four phases we see most often inside mid-market product and engineering organizations. Where you sit depends on what you have built around the tools, not how long you have been at it. Score the questions below and you will see the pattern you are in and what has to change next.

The four phases we see most often

Phase 1

Shadow and scattered AI

Individuals are experimenting. Learning is private. Leadership has low visibility.

Phase 2

Tool rollout

The company has bought access and created visible activity, but wins are still scattered and fragile.

Phase 3

Workflow infusion

AI is starting to improve recurring work across the PDLC. Gains begin to survive the handoffs.

Phase 4

System shift

The company is changing decisions, intake, measurement, and operating rhythms, not just workflows.

→

AI Catalyst helps teams move from tool rollout to workflow infusion faster, and helps teams further along the journey strengthen what is already starting to work.

See where the gains are breaking down

Score each question from 1 to 5 based on what is most true in your organization today. You'll see what phase of the journey you're in, what pattern is holding you back, and what has to change next.

1. Functional spread of AI usage
Across product, design, engineering, and QA, how broadly has AI spread into day-to-day work?
This is about how far AI has spread across product, design, engineering, and QA, not whether every individual is using it.
1 AI use is isolated to a few individuals or experiments.
2 AI is showing up in one function, but barely outside it.
3 AI is active in two functions, but not broadly across the PDLC.
4 AI is active across most major functions, though unevenly.
5 AI is active across product, design, engineering, and QA as part of normal work.
2. Consistency of usage and output
Within the functions where AI is already present, how consistent is usage and output quality across teams and individuals?
This is not about how many functions are using AI. It is about whether usage is consistent enough inside those functions to create reliable value instead of hidden rework.
1 Very inconsistent. A few standout people drive most of the value.
2 Uneven. Some teams are using AI well, others barely use it or get weak results.
3 Mixed. There are pockets of consistency, but outcomes still depend heavily on who is doing the work.
4 Mostly consistent. Most teams have a workable baseline, though some variation remains.
5 Consistent. Capability is broad enough that results do not depend heavily on a few standout individuals.
3. Shared workflows
How many recurring workflows have shared AI practices, not just individual use?
1 None. AI is mostly personal and ad hoc.
2 One or two workflows have informal shared patterns.
3 A few recurring workflows have shared templates, expectations, or review norms.
4 Several important workflows have shared practices across teams.
5 Shared AI workflows are established across multiple recurring parts of the PDLC.
4. Handoff survival
When AI helps one team move faster, how often does that gain survive the handoff to the next team?
1 Almost never. The gain dies in the transition.
2 Rarely. Some speed is created, but the next team usually has to redo or reinterpret the work.
3 Sometimes. Some gains survive, but handoffs are still inconsistent.
4 Often. Handoffs usually preserve the value created upstream.
5 Consistently. AI gains carry through transitions and improve end-to-end flow.
5. Measurable outcomes
Can you point to clear PDLC outcomes that improved because of AI?
Think less rework, faster review cycles, better handoffs, fewer defects, faster release readiness, or stronger requirement quality.
1 No clear outcomes. Mostly activity and anecdotes.
2 Weak evidence. Some local wins, but no strong proof.
3 A few visible improvements, but not yet repeatable or broadly trusted.
4 Clear improvements in one or more PDLC outcomes.
5 Clear, repeatable improvements in multiple PDLC outcomes tied to delivery or business impact.
0 / 5 answered
// Your Results

Your phase and pattern

What's happening now

What's at risk

What needs to change next

If the gains are real, but they are not surviving the system, that is the work.

This is where AI Catalyst helps teams shorten time to value.

Fix where AI value is dying See how to turn uneven gains into shared workflows
Research behind this diagnostic

McKinsey, The State of AI: How Organizations Are Rewiring to Capture Value (2025): workflow redesign had the biggest effect on whether organizations saw EBIT impact from gen AI.

Faros AI, The AI Productivity Paradox Research Report: 75% of engineers use AI tools — yet most organizations see no measurable performance gains. High-AI-adoption teams complete 21% more tasks and merge 98% more PRs, but PR review time rises 91% — the bottleneck absorbs the gain.

DORA / Google Cloud, State of AI-assisted Software Development 2025: 90% of developers use AI; 80% perceive productivity gains; epics per developer up 66.2%. AI doesn't fix a team; it amplifies what's already there. The greatest return comes from internal platforms, workflow clarity, and team alignment — not the tools themselves.

Workday, AI Friction Into Flow (2026): for every 10 hours of productivity gained, about 4 hours were paid back in rework, correcting, clarifying, and refining AI output.

Section, AI Proficiency Report (2026): significant perception gaps exist between leadership and individual contributors in whether AI is saving time and how people feel about it.