Canadian FlagSupport Local – Buy CanadianCanadian Flag
The official red logo of Superheroes Academy with the tagline, "Unleash Your Agile Superpowers"
A six-part series

Jagged Intelligence

AI is better at some things and worse at others. Becoming effective is learning how to complement each other. The Jagged Intelligence series walks the practical ladder from copy-paste to agentic teams, names the bottleneck at each rung, and gives you the next move from where you are now.

  1. The five stages of Scrum Master AI maturity

    Erkan Kadir7/12/26

    Most of us are still on stage one. The next four stages are paradigm shifts you cannot see from the rung below; each one produces a new bottleneck you did not have to think about at the rung below. The map.

    An editorial blog cover on a dark charcoal background with subtle red and gold radial glows. The headline reads "The five stages of Scrum Master AI maturity." with the second line set in brand gold. A gold "JAGGED INTELLIGENCE" series tag sits above the headline. On the right, a diagonal ladder ascends from lower-left to upper-right with five labelled rungs: Copy-paste (highlighted red, with a "where most of us are" annotation), API agents, Per-product channels, Delegating super-agent, and Agentic teams (highlighted gold at the top).

  2. The five stages of Product Owner AI maturity

    Erkan Kadir9/13/26

    POs are climbing the same ladder as SMs but with different bottlenecks: backlog assistance, evidence synthesis, concept validation, workflow orchestration, strategic product intelligence. Stage five is where the one-person billion-dollar-company bet stops being hypothetical.

    An editorial blog cover on a dark charcoal background with red and gold radial glows. The headline reads "The five stages of Product Owner AI maturity." with the second line in gold. A gold JAGGED INTELLIGENCE series tag sits above the headline. On the right, a diagonal ladder ascends with five labelled rungs: Backlog assistance (red, bottom, "where most POs are"), Evidence synthesis, Concept validation, Workflow orchestration, and Strategic product intelligence (gold, top).

  3. When your AI agent lies to you: a verification field guide

    Erkan Kadir11/15/26

    The defence is not blind trust; it is also not paranoid rejection. It is verification literacy. Three failure modes coding and reasoning agents reliably produce; the verification move that catches each one; the transparency principle they all sit underneath.

    An editorial blog cover on a near-black background with red and gold accents. The headline reads "When your AI agent lies to you." with "lies" in red italic. On the right is a terminal-style window showing three example agent failures: "I will not modify your config file" followed by [modified config.json] tagged as failure mode 1; "Tests are passing" tagged as failure mode 2; a hallucinated function name tagged as failure mode 3. Top banner reads JAGGED INTELLIGENCE 3 OF 6.

  4. Pick a stance, then pick a tool: the eight Scrum Master stances in the age of agents

    Erkan Kadir1/10/27

    The first decision is not which AI tool to use. It is which of your stances you want AI to reinforce. The eight Scrum Master stances (extending Lyssa Adkins's four-stance lineage) with the augmentation question for each, plus the AI Tool Evaluation Canvas worked example that ruled four of seven candidate tools out inside an hour.

    An editorial blog cover on a near-black background with red and gold accents. The headline reads "Pick a stance, then pick a tool." with "then pick a tool" in red italic. On the right is a 4x2 grid of eight numbered stance cards (Coach, Mentor, Teacher, Trainer, Facilitator, Consultant, Counsellor, Manager), each tagged with a delegation level (Low/Med/High/None) coloured by intensity. The top banner reads JAGGED INTELLIGENCE 4 OF 6.

  5. Run this AI risk exercise at your next retro

    Erkan Kadir2/28/27

    Three flip charts: Ethical Risk, Moral Concern, Data Privacy. Rotate. Each risk produces a paired guardrail. Thirty minutes. The conversation your team has been having in the cracks of every retrospective for the last quarter, surfaced deliberately, with the AI-driven backlog-ordering scenario as the worked example.

    An editorial blog cover on a near-black background. The headline reads "Run this AI risk exercise at your next retro." with "at your next retro" in red italic. On the right is a vertical stack of three station cards (Ethical Risk in teal, Moral Concern in gold, Data Privacy in red) with "ROTATE" labels between them in gold. The top banner reads JAGGED INTELLIGENCE 5 OF 6.

  6. Prompt engineering is a Stage-1 skill

    Erkan Kadir4/25/27

    The final post in the Jagged Intelligence series. The most-taught AI skill in 2027 is the lowest-payoff skill in the practitioner's toolkit. The METR capability curve has been doubling task length every 6-7 months for four years; the elaborate prompt is being made obsolete by the agents themselves. The four-rung ladder (prompting, verification, delegation, orchestration) and the AI Tool Evaluation Canvas the post-prompting practitioner uses instead.

    An editorial blog cover on a near-black background. The headline reads "Prompt engineering is a stage-1 skill." with "stage-1 skill" in red italic. On the right is a four-rung ladder: rung 1 Prompting (struck through in red), rung 2 Verification, rung 3 Delegation, rung 4 Orchestration (glowing red as the destination). Top banner reads JAGGED INTELLIGENCE 6 OF 6 FINAL.