Building Coach Bob’s Playbook – Personas in Play

Rivalries make football memorable — and they reminded me why personas, whether in AI or research, are about perspective, not perfection.

Rivalries & Passion

There’s an old Woody Hayes quote — maybe apocryphal, but legendary all the same. After going for two points late in a lopsided game, the Ohio State coach was asked why. His reply: “Because I couldn’t go for three.”

Real or not, it’s become part of football lore. And that’s the thing about rivalries: they create passion, emotion, and meaning.

In Coach Bob’s world, rivals served the same purpose. And in research, personas do too.

Meet the Rivals

Once Coach Bob had his roster of synthetic players, it was time to put him on the field against competition.

Bob himself became the “balanced” coach — the one grounded in consistency. To give him contrast, I extended the same process into two new models:

• Coach Sal: a run-heavy strategist.
• Air Raid Rob: a pass-first, throw-it-deep risk-taker.

They weren’t designed to perfectly copy real coaches. They were built to represent styles — distinct perspectives that would shape play calling differently and make the game feel alive.

Personas, Not Replicas

That’s the key: rivals weren’t replicas, they were personas. Just like Bob.

Even simple changes could shift how they came across. A small tweak — like telling a persona “you are angry” — was enough to change tone, cadence, and perceived intent. That nuance made the differences feel real, even if they weren’t perfectly human.

But minor wording tweaks were never enough. Bob’s personality was always the product of design. From the start, it took deliberate framing, prompt iteration, and few-shot examples to make him sound natural. Building Bob provided the process template and lightened the load, but rivals required the same kind of care, just with different emphases.

The lesson: personas don’t need to be exact copies. They just need to be consistent perspectives that reflect the style they are meant to represent.

The Research Parallel

In market research, personas play the same role. They’re not digital twins designed for exact replication — they’re designed to reflect the motivations, tendencies, and perspectives of a segment.

And while GenAI makes it easy to say “act like…” and get something back, that’s only the first step:

• Basic Prompting: “Act like… and answer this question” → easy, but shallow and reflects only general LLM knowledge.
• Contextual Prompting: “Here’s a brief profile of a real customer — respond as if you are them” → better, but untested.
• Validated Digital Asset: Built from real human data, carefully constructed with rigor → that’s where the value lies.

Much like Bob and friends required careful testing and a process template for success, developing digital assets for research that truly impact outcomes requires putting in the work. Building the playbook of what works, what doesn’t, and the line between useful and misleading is where the real differentiation and value lies.

Closing Thought

Rivalries drive passion in football. Contrasting personas drive insight in AI and research.

Bob’s rivals reminded me of the same truth my day job reinforces: with the power of LLMs today, anyone can build a digital persona — but excellence comes from the rigor it takes to make it meaningful.