Using PDSA to keep AI honest

AI is brilliantly persuasive, but it’s not always right.

Artificial Intelligence is changing how we learn, lead and improve. It’s fast, fluent and, let’s be honest, often very convincing. That’s the upside. The risk? Its confidence can mask inaccuracies, especially when we’re up against a deadline. But PDSA and Agile sprints give teams the structure to learn quickly without being misled.

PMI’s Anthony Richardson and Susannah Clarke recently wrote an article for the Chartered Quality Institute exploring the simple antidote of combining rigour with speed. They demonstrate how the PDSA cycle (Plan, Do, Study, Act) pairs beautifully with Agile sprints to help teams test AI tools safely, surface errors early, and turn learning into better outcomes.

Look out for these key insights:

• The encouragement trap: AI’s tone can make results feel “right”. Structured feedback loops keep us curious and critical.

• Small experiments, big gains: Short sprints (think half‑day Kaizen‑style sessions) let cross‑functional teams compare outputs, refine prompts, and agree whether to adopt, adapt or abandon.

• Leadership matters: Create psychological safety, protect time for experimentation, and celebrate learning, not just results.

For those of us trialling AI in our work, this piece is a practical guide to doing it quickly and well. Read the full CQI article now, and begin building systems that improve smarter.

“Failing fast is not about failing often, it’s about learning quickly and adapting wisely.” — Randy Bean