Trusting the Number: Respect for People, AI Literacy, and KPIs at July's Lean Coffee

By the time Central Coast Lean’s July Lean Coffee got underway, there were almost as many AI notetakers logged into the call as there were people. That felt about right for the morning ahead. For the next seventy minutes, whether the topic was people programs, AI adoption, or metrics, the group kept circling one question: what does it take to trust something — a machine, a metric, a claim of impact — enough to act on it?

Balance between evidence and insights

Central Coast Lean’s Lean Coffee runs without a set agenda. Practitioners show up, pitch topics in about thirty seconds, vote with the Lean Coffee Table app, and talk through whatever rises to the top. Seven people signed in this month, among them lean consultants, a practitioner working in healthcare, a continuous-improvement professional, and a software engineer who builds AI tools scoped to particular jobs. They spent the session on questions that will sound familiar right now: how do we show that investing in people pays off, how do we build AI capability with no single blueprint for it, how do we measure the right thing instead of the easy thing, and how do we let people experiment with AI without losing track of what is being built.

When Respect for People Meets the Spreadsheet

This drew the most votes of the day. Mark de Kiewiet has been telling clients for years that they do not have the money not to invest in Respect for People. His strongest evidence right now comes from a bakery: weekly HR complaints there fell from six or eight to zero, and an operations manager who had spent years patching the same recurring problems is finally watching root causes get solved. What Mark does not have yet is a dollar figure.

Ravi David pushed the group to be careful with that gap. Document what changed and what was observed, he suggested, rather than assuming a straight line between the two.

Janice Hodge offered a harder-edged caution from her own corporate past, where people programs sometimes existed mainly to look good in a shareholder report, disconnected from anything frontline employees felt. Katelyn Carey, who works in healthcare, said the opposite problem shows up in her world. Not much moves without numbers, so she builds a chain from leader behavior to turnover to incivility to patient outcomes just to get a program funded.

From Data Literacy to AI Literacy — Three Different Roads In

Eric Olsen brought this topic in after a June 10, 2026 webinar hosted by The Ohio State University Center for Operational Excellence, where a practitioner from Worthington Steel described building data literacy first and AI literacy on top of it. What struck the group was how differently that same journey looks in practice.

Janice Hodge described a former employer’s crowdsourced model, where volunteers were trained centrally and then took the training back to their own teams, tailoring the examples as they went. That beat generic demos that do not connect to anyone’s actual job. Josh, who builds AI systems professionally, goes the opposite direction: rather than one interface for everyone, his team builds tools scoped to specific roles, and keeps some steps deterministic instead of handing them to a model just because it is possible.

Katelyn Carey pointed out that even sharing a ready-made prompt is a teaching tool, since it shows people how much specificity changes the answer they get back.

Measuring What Matters: Goal KPIs vs. Execution KPIs

Eric has been working through this with his Cal Poly students. It is one thing to say you should measure attendance, another to operationalize that measurement so it means something. His current approach is to write quizzes that require having been in the room and paying attention, rather than taking attendance at all.

Mark de Kiewiet offered a cleaner distinction: a goal KPI is the outcome you actually want, an execution KPI is a means of getting there, and a simple cause-and-effect diagram connects them. Olga Zhuravel grounded the conversation in a habit she returns to constantly, which is to ask what the problem statement and objective are before ever picking a metric. It is tempting to reach for whatever data is easiest to get, she noted, whether or not it is the data that matters. Used that way, a measure quietly turns into a weapon instead of a guide.

Playground or Production? Governing AI Experimentation

As the formal agenda wrapped up, Olga Zhuravel raised a question that clearly was not finished being asked. After telling employees to go play with AI, how do you build governance around it once everyone is building and no one is sure who is actually upskilling?

Josh laid out the approach his team is settling into. Keep experimentation open, with guardrails like internal traffic monitoring to catch exposed personal information, but route anything moving from one person’s own use toward team-wide deployment through a more rigorous process. He compared it to interior decorating versus hiring a general contractor: decorate your own room however you like, but once you are touching load-bearing walls, you want the contractor. Eric noted that the metaphor traces back to a Central Coast Lean post from last month on remodeling a home without a general contractor (purpose-ccl.org/post/building-ai-without-a-general-contractor), so Josh’s comment extended an idea already in play rather than inventing one. Ravi David offered the long view, comparing the current AI moment to the early days of personal computers, when companies tried and largely failed to control adoption tightly. It took something like five to ten years, he noted, before the real value came through anyway.

What seems to be emerging, at least from this one conversation, is the same tension wearing three different outfits. Respect for People needs evidence without turning into a spreadsheet exercise. AI adoption needs structure without collapsing into one-size-fits-all. KPIs need to be easy to track without drifting from the goal they were meant to serve. In our experience so far, the thread running through all three is discipline about the objective: asking what problem we are actually solving before reaching for the nearest available number or the newest available tool. We are still learning our way through this, and we are curious how others are handling it, which is rather the point of a Lean Coffee.

Continue the conversation. Central Coast Lean’s Lean Coffee meets virtually on the second Wednesday of every month at 10:00 a.m. Pacific, and it is free. Details and the Lean Coffee Table link are at purpose-ccl.org/lean-coffee. Bring a topic, or just bring your coffee, and help us make the work better, faster, and easier.

Knowledge Map

Process Keywords: Lean Coffee, Respect for People, root cause analysis, cause-and-effect diagram, goal KPI, execution KPI, operationalizing a metric, crowdsourced peer-led training, data literacy, AI literacy, role-scoped AI tools, deterministic process steps, prompt sharing, internal traffic monitoring

Context Keywords: no dollar figure for people work, unproven AI return, metric gaming, generic AI training that misses the job, ungoverned AI building, disconnected people programs, shareholder-report programs, frontline distrust of leadership metrics

Application Triggers:

• If your Respect for People evidence is only anecdotal, then document what changed and what was observed before claiming causation.

• If your organization has data literacy but not AI literacy, then try role-based, peer-led training instead of generic demos.

• If a prototype is moving from individual use to team-wide deployment, then route it through a review before scaling.

• If a KPI is easy to measure but does not map to your objective, then treat it as a proxy, not the goal.

• If you cannot connect a people program to a metric leadership already tracks, then expect it to be funded as a nice-to-have.

Related Continuous-Improvement Themes: voice of the process versus voice of the customer, standard work for emerging technology, psychological safety, servant leadership, kaizen applied to AI adoption, measurement systems that do not distort behavior

This post reflects a July 8, 2026 Lean Coffee conversation hosted by Central Coast Lean, synthesized with Claude AI assistance and reviewed by the practitioners named in it. Corrections welcome.

Eric O. Olsen, PhD — Director, Central Coast Lean; Professor of Industrial Technology, Cal Poly San Luis Obispo. eric.o@centralcoastlean.org | purpose-ccl.org