Top 7 AI champion discovery examples for spotting power users in teams

Quick answer: AI champion discovery is the process of identifying the people already using AI in ways that are repeatable, useful to others, and tied to real work outcomes—not just the loudest volunteers or the heaviest chat users. The best examples come from observed [[workflow](/how-to-onboard-new-hires-to-your-team-ai-workflow/)](/ai-workflow-redesign-audit-[checklist](/what-leaders-need-to-see-checklist/)/) behaviour: people who turn one-off prompting into reusable assets, unblock peers, navigate governance constraints, and show evidence of better output. If you want champions who can actually raise [[adoption](/shallow-ai-adoption-in-marketing-teams/)](/ai-adoption-briefing-workflow-best-[[[[[practices](/skills-interview-for-ai-roles-best-practices/)](/legal-workflow-automation-with-ai/)](/german-works-council-ai-approvals/)](/ai-[rollout](/ai-rollout-not-sticking-task-decomposition/)-blockers-best-practices-2026/)](/ai-champion-[enablement](/long-term-ai-enablement/)-best-practices-2026/)/), look for proof in work patterns, artifacts, and peer pull—not self-reported enthusiasm.

TL;DR

• Most teams have some AI [usage](/real-ai-usage-data-examples/) but very few true power users; broad access does not equal workflow change.
• The strongest champion signals are behavioural: reusable [workflows](/ai-workflows-for-marketers/), peer teaching, output judgment, and visible problem-solving under real constraints.
• Good discovery combines three evidence types: what people say, what they can show, and what peers confirm.
• The wrong method is common: picking volunteers, title holders, or prompt hobbyists who cannot transfer their practice to the team.

What does good AI champion discovery actually look like?

If you have already rolled out ChatGPT Enterprise, Copilot, Gemini, Claude, or internal AI tools, you have probably seen the pattern: lots of logins, lots of curiosity, and not much durable workflow change (Survey: How Executives Are Thinking About AI in 2026). That gap is not unusual. BCG reported that 60% of companies were not generating material value from AI despite substantial investment, and that more than 85% of employees still sat in mid-level adoption stages while fewer than 10% had reached more advanced use (AI Adoption Puzzle: Why Usage Is Up But Impact Is Not | BCG).

That is why AI champion discovery matters. You are not trying to find the people who “like AI.” You are trying to find the people who can help a team move from access to adoption.

In practice, a strong AI champion usually does four things:

1. Uses AI on real recurring tasks, not just experiments.
2. Produces outputs others can inspect or reuse.
3. Knows where AI fails and applies judgment.
4. Helps peers get unstuck after rollout, not just during launch.

This is also why surveys are weak on their own. People overstate usage, confuse experimentation with impact, or rate themselves highly because they know the vocabulary (AI in the workplace: A report for 2025 | McKinsey). McKinsey’s workplace AI research also points to a practical pattern: managers in the 35–44 range often self-report high enthusiasm and experience, which can make them natural change agents—but that is still only a starting signal, not proof of champion behaviour.

The examples below are the signals worth trusting.

The top 7 AI champion discovery examples

1. The person who turned a one-off prompt into a repeatable workflow

This is the clearest signal. A power user stops asking AI random questions and starts building a repeatable sequence around a recurring task: draft, critique, revise, verify, export, handoff. That might be a recruiter who uses AI to turn intake notes into structured scorecards, or a marketer who converts webinar transcripts into campaign variants with a review step.

The key is repeatability. OpenAI’s guidance on internal champions makes the same point: the strongest use cases are the ones another person or team can follow and reproduce, not a long list of isolated experiments (Turn AI use cases into visible impact - Resource | OpenAI Academy).

What to look for: - A named workflow with clear steps - Inputs and outputs that can be shown - A saved prompt, template, or SOP - Evidence that someone else has copied it successfully

Concrete example: in a finance team, one analyst uses AI to classify vendor spend anomalies, draft explanations, and prepare a first-pass monthly variance note. The champion signal is not “uses AI daily.” It is “reduced a 90-minute recurring task to 25 minutes with a documented review checklist” (BCG AI Radar 2026: As AI Investments Surge, CEOs Take the Lead).

2. The teammate others already ask for help

Peer pull is a better signal than self-nomination. If three people independently say, “Ask Nina, she figured out how to use Copilot for client summaries,” that matters more than Nina volunteering for the AI committee.

GitHub’s playbook for internal AI champions describes why this works: adoption spreads faster when people see a trusted teammate share a specific prompt or workflow that saved real time on real work.

What to look for: - Repeated mentions in peer interviews - Slack or Teams history where colleagues ask for help - Informal office hours, Looms, or internal docs created without being asked - Cross-functional reach, not just expertise inside one niche

Concrete example: in HR, one business partner becomes the default person for writing better job descriptions, interview rubrics, and policy drafts with AI. She is not the most technical person in the team. She is the one whose examples others can actually use.

This matters because champions are multipliers. A solo expert who cannot transfer practice is useful, but not enough. A discoverable champion creates local proof.

3. The person who shows artifacts, not just confidence

A lot of false positives sound impressive in conversation. They know the terms, talk about agents, mention prompt [engineering](/vp-engineering-ai-rollout/), and say they use AI “all the time.” Then you ask what changed in their workflow, and there is nothing inspectable.

Real champions can show artifacts: - Prompt libraries - Before/after versions of deliverables - Evaluation rubrics - AI-assisted research notes - Templates used by others - Decision logs showing where AI was accepted, edited, or rejected

This matters because sophisticated use is visible in the work. Research discussed in Forbes highlighted markers such as specific prompting strategies, ambitious requests, and comfort with the tools as signs of more advanced use (Why We Don’t Have More AI Power Users In The Age Of AI).

Concrete example: a legal ops manager shows a clause-comparison workflow with three saved prompts, a red-flag checklist, and examples of where the model hallucinated a citation. That is champion material. It shows both usage and judgment.

If someone cannot show anything beyond screenshots of chats, be careful. Heavy usage without artifacts often means curiosity, not transferable capability.

4. The person who uses AI for harder judgment tasks, not just drafting

Basic drafting is now common. Power users go further. They use AI to compare options, identify patterns, challenge assumptions, and improve decisions. Forbes gave a useful example: instead of just drafting emails, a power user might ask AI to analyse why a product feature is not gaining adoption by reviewing user feedback and surfacing patterns.

This is one of the best discovery signals because it separates surface use from deeper workflow integration.

What to look for: - AI used in diagnosis, prioritisation, QA, or scenario testing - Prompts that ask for critique, alternatives, or edge cases - Evidence of human review and final judgment - Use across multiple steps of a task, not just first draft generation

Concrete example: a customer success lead feeds churn reasons, support tickets, and NPS comments into an AI workflow to cluster themes, draft interventions, and stress-test the proposed action plan. The champion signal is not speed alone. It is better judgment under time pressure.

Inc’s description of power users fits here too: they treat AI like a junior colleague, iterate instead of giving up after one weak answer, and build reusable workflows rather than one-off queries.

5. The person who works within governance instead of around it

In many EU teams, the practical blocker is not interest. It is uncertainty: what data is allowed, which tools are approved, whether works councils need involvement, what legal review is required, and how to handle customer information (The State of Organizations 2026).

A real champion does not ignore those constraints. They learn how to work inside them. That makes them far more valuable than the “shadow AI” enthusiast who gets results by bypassing policy.

What to look for: - Uses approved tools and knows their limits - Can explain safe vs unsafe use cases in plain language - Escalates governance questions early - Helps translate policy into workable team habits

OpenAI’s champion-role guidance is practical here: strong champions notice access issues, governance questions, and workflow blockers, then route them to the right owners.

Concrete example: an operations manager wants to use AI for meeting summaries and SOP drafting. Instead of uploading sensitive customer data into a public tool, she works with IT and legal to define a safe pattern using approved enterprise tooling and redacted inputs. That person is much more useful to the business than the employee who gets flashy results in an unapproved app.

6. The person whose examples spread across the team

A champion is not just a high performer. A champion creates adoption spillover. You can see it when one person’s workflow becomes a team habit.

This is where discovery should move beyond individual skill and ask: did anyone else adopt their method? If yes, how far did it spread? McKinsey’s broader organisational research points to a small group of AI Pioneers pulling ahead, which reinforces the idea that leading teams distinguish themselves not by access alone but by embedding AI into how work gets done.

What to look for: - Shared templates or prompt packs in active use - Team norms changed because of one person’s example - New joiners onboarded into the workflow - Measurable reduction in repeated questions after the workflow was shared

Concrete example: in a sales team, one rep builds a call-prep workflow using CRM notes, account history, and public research. Within six weeks, four other reps adopt the same structure. Pipeline reviews become more consistent. That is a stronger champion signal than one rep quietly outperforming with a private setup.

The practical test: can this person create second-order adoption?

7. The person who keeps improving after the first launch

Many teams mistake early excitement for champion potential. The better signal is persistence after the novelty wears off.

OpenAI’s champion guidance says strong champions pay attention to what happens after launch or training: where people succeed, where they get stuck, what support is missing, and how workflows need to evolve.

This matters because most AI rollouts stall after the first wave. BCG’s 2026 AI Radar also argues that leadership capability-building matters; CEOs who spend at least eight hours a week building their AI capabilities are more likely to generate meaningful value. The same pattern holds lower down: sustained learning beats initial enthusiasm.

What to look for: - Updates prompts and workflows over time - Tracks failure modes and fixes them - Revises examples after model or policy changes - Continues helping peers months after rollout

Concrete example: an engineering manager first shares a code review prompt pack, then notices developers struggle with context windows and noisy diffs. She updates the workflow, adds repository-specific instructions, and creates a short “when not to use this” guide. That is champion behaviour because it compounds.

How to discover champions without relying on surveys

If you want a reliable shortlist, use a simple three-layer evidence model:

1. Interview evidence Ask people to describe one recurring task where AI changed the way they work. Push for specifics: inputs, prompts, review steps, failure cases, and outputs. Generic answers usually collapse under follow-up.

2. Artifact evidence Ask them to show the workflow: prompt packs, documents, templates, recordings, examples, or metrics. This is where false positives drop away fast.

3. Peer evidence Ask managers and teammates who others go to for help, whose examples spread, and who made AI useful in day-to-day work.

This works better than surveys because it captures behaviour, not identity. It also avoids the common trap of over-selecting extroverts, senior people, or AI hobbyists.

A practical scoring lens is: - Observed: we heard a credible workflow description - Verified: we saw artifacts or outputs - Confirmed: peers or managers independently pointed to the same person

That is enough to rank likely champions without turning discovery into a six-month programme.

One more point: do not only search in technical teams. Some of the best champions sit in HR, operations, support, finance, and marketing because they face repetitive language-heavy workflows where AI can quickly become useful. In many companies, these teams also have more change-management friction, so a credible internal champion matters even more.

Quick answer: A simple AI champion discovery playbook

If you need to run discovery in a real company, keep it tight: one owner, two weeks, one shared scorecard. In most companies, the best owner is L&D, transformation, HR, or the internal AI lead, with team managers supplying nominees and context. Start with 10–20 candidates across functions, using three inputs: usage analytics as a lead list, manager nominations, and peer mentions. Then run 20–30 minute interviews and ask the same five questions every time:

1. What recurring task do you now do differently with AI?
2. Can you walk me through the steps, including review and failure checks?
3. What artifact can you show: prompt, template, SOP, before/after output, or metric?
4. Who else has adopted this workflow or asked you for help?
5. What changed in speed, quality, consistency, or volume?

Score each candidate 0–2 on four dimensions: repeatable workflow, artifact proof, peer pull, and judgment/governance. A simple shortlist rule works well: prioritise people scoring 6+ out of 8, with at least one artifact and one peer confirmation. To compare across teams, rank by total score first, then by spread potential: can others copy this quickly?

After identification, do not stop at the list. Give each champion one immediate job for the next 30 days: document one workflow, run one peer session, and log blockers. Track proof of impact through reuse metrics, attendance, copied templates, reduced task time, or manager-reported workflow adoption.

Common mistakes that produce the wrong champions

The most common mistake is choosing volunteers. Volunteers are useful for momentum, but they are not automatically the people others trust or the people with the strongest workflows.

Other bad selection patterns:

• Picking by title The Head of AI, innovation lead, or team manager may be important sponsors, but not the best day-to-day champions.

• Picking by tool usage volume Heavy usage can mean lots of low-value chat activity. It does not prove workflow change or good judgment.

• Picking the most technical person Some of the best champions are translators. They make AI usable for normal work, not just advanced demos.

• Ignoring governance behaviour If someone gets results by bypassing policy, they may create risk faster than adoption.

• Confusing charisma with transferability A compelling presenter who cannot show reusable assets is not yet a champion.

• Treating discovery as a one-time event Champions change. New ones emerge after workshops, hackathons, or tool changes. Re-measure quarterly if AI adoption is a serious priority.

This is where many enablement efforts fail. They build a champions network around enthusiasm instead of evidence. Then six months later, the “champions” are inactive, peers are unconvinced, and leadership concludes the team “just isn’t ready.” Usually the issue was selection, not readiness (Playbook series: Activating your internal AI champions · GitHub).

Bottom line

If you want to spot AI power users in teams, do not ask who is excited about AI. Ask who changed a real workflow, can show the evidence, and helped someone else do the same. The seven best discovery examples are all behavioural signals: repeatable workflows, peer pull, visible artifacts, better judgment, governance fluency, spread across the team, and post-launch persistence.

That is the practical threshold. Not “who talks about AI most,” but “who makes AI useful in a way others can copy. ” If you identify champions that way.