Video summary

After This Video, You Will Actually Understand AI Enablement | #PowerTalks Ep. 91

Main summary

Key takeaways

Business

AI enablement thesis (why pilots fail to deliver value)

AI adoption feels stuck because organizations try to “apply AI” to existing workflows without fixing the underlying process, ownership, and data/governance issues.

Core blocker: legacy process debt + missing governance + unclear ownership, which leads to pilots that never scale or prove measurable ROI.

Key frameworks / playbooks mentioned

“AI Turning Point” (ebook; process-first transformation narrative)

  • Focuses on organizational readiness
  • Encourages teams to:
    • Pause
    • Identify what’s broken
    • Fix before building/expanding AI

SmartStart service / phased rollout playbook (ProArch)

A phased approach designed to reduce change-management failure and delays:

  • Establish cadence and prioritization for what to address first
  • Start with small pilot groups
  • Target high-value, low-effort use cases to generate early ROI
  • Avoid getting stuck in only “whiz-bang” high-production work before early wins
  • Avoid the “big bang approach” (all-in rollout), since it often fails due to change-management issues

Center of Excellence (CoE) + Champions model

  • Build an AI Center of Excellence (CoE)
  • Assign champions/owners for ongoing responsibility and accountability

Quick wins via persona/task mapping + “3 D’s”

A workshop format to find practical opportunities:

  • Map:
    • Personas/roles → top tasks
    • Tasks that are dull, distracting, draining (the “3 D’s”)
  • Use a priority matrix/workbook to rank candidate use cases using criteria-based selection

Process mapping before agent building

For each automation/agent:

  • Map the existing process
  • Confirm dependencies and look for hidden inefficiencies
  • Simplify the process before adding AI
  • Test in chat first (iterative response shaping) before implementing as a full agent

Concrete examples / case illustrations

  • Approval workflow inefficiency

    • Pattern: a process with a two-stage approval where no one can explain why
    • Recommendation: revisit and simplify the workflow before applying AI

  • Agent development: data/file dependency correction

    • Teams describe a workflow (“go here, do this, do that”), but lack specifics about where the needed data lives
    • Approach: use chat-based prototyping to discover missing files/data and restructure the workflow to reduce complexity

  • Transparency and bias in hiring (SaaS HR screening)

    • A company used AI to screen candidates without transparent disclosure to subscribers
    • Result: discovered bias led to a major lawsuit triggered via a California FOIA request
    • Lesson: transparency helps prevent downstream harm and supports faster correction

How to find “quick wins” (actionable recommendations)

  • Run a cross-functional workshop to:
    • Identify role personas and their 5–6 core tasks
    • Capture pain points using the 3 D’s (dull/distracting/draining)
    • Translate pain into AI use cases (e.g., no-code agents for repetitive tasks)
  • Use a priority matrix to select use cases that:
    • maximize early impact
    • reduce implementation effort
  • Build early wins that demonstrate measurable business outcomes before scaling

Governance & accountability requirements (execution guidance)

  • Governance is repeatedly emphasized alongside use cases:
    • establish guardrails
    • define review/approval workflows
    • implement data governance
  • Ownership cannot be “part-time.”
    • Don’t assign responsibility like “tap Terry in finance to be the AI person”
    • Ensure dedicated owners/champions/CoE accountability
  • Human-in-the-loop as a practical risk-control mechanism:
    • AI can draft/read/score
    • critical decisions should retain human review

Standards / risk boundaries (high level)

  • Use organizational values and ethics to guide deployment
  • Avoid sensitive categories of use cases that can:
    • deny consequential services
    • create risk of harm (e.g., medical prescribing without appropriate controls)
    • infringe human rights
  • Safer alternative framing:
    • Use AI to support eligibility/approvals rather than deny outcomes
    • Example: help justify insurance claims or complete documentation to access services

Notable metrics & KPIs cited

  • AI adoption scaling failure
    • 80% of AI pilots fail to scale into production and deliver measurable business value
  • General project success benchmark (context)
    • 60% of IT projects finish on time and on budget
    • 20% finish within 50% over budget
    • 20% finish more than 50% over budget
  • These are used to argue against overly penalizing pilot overruns without accounting for:
    • change-management realities
    • the need for phased execution

Presenters / sources

  • Griffin Lickfeld (host)
  • Jim Spignardo (Director of Cloud Strategy & AI Enablement; ProArch; author of The AI Turning Point)
  • Mentions:
    • Microsoft AI principles (referenced)
    • Omar El Takouri (podcast: The Department)

Original video