← narwal.one/Second Brain
SecondBrain
Ask the Brain
Index/Sourceupdated Sun Sep 27 2026 08:00:00 GMT+0800 (Philippine Standard Time)

How AI Is Changing Talent Not Just Tasks (HBR IdeaCast)

hbrideacasttalent-strategyhuman-judgmenthuman-at-the-helmrole-redesignentry-levelinternal-mobilitysalesforcegoldmanhr

How AI Is Changing Talent Not Just Tasks (HBR IdeaCast)

type: source — bonus episode of HBR IdeaCast, the first of its Thursday series on how AI is changing what it means to be an executive. Host Adi Ignatius (HBR editor at large; co-host Alison Beard introduces) interviews Paula Goldman, Salesforce's chief ethical and humane use officer, US-government AI-policy adviser, and author of Manage the Machine: How to Harness Human-AI Collaboration at Work. Full HBR transcript captured via Web Clipper.

Central move: the AI-and-work debate is stuck on labour-market policy (how many jobs) and productivity (how much cheaper). Goldman argues the neglected lever is work design — AI takes tasks, not roles, so the remaining parts of every role get more important, and leaders must deliberately decide where human judgment carries the day.

Key claims

  1. Human at the Helm, not human in the loop. "Human in the loop" is a Cold-War military term (who makes the consequential call when radar detects a missile?). In the AI wave it degraded into "AI drafts, people approve" — which "does not really work very well" for agents that reason through many complex tasks at speed. Instead: put the right things in front of human judgment at the right time. She offers it as both a system-design principle and a management metaphor. See Human in the Loop.
  2. "Routine to AI, complex to humans" gets you ~70% of the way. The remaining 30% is customer emotions and preferences, employee emotions and preferences, and sensitive topics that "only people could handle even if AI can". Her rule of thumb for service: angry customers want a person; embarrassed customers want the AI. The healthcare example: a person always schedules the first appointment after a new cancer diagnosis — not a capability gap but "something to preserve". See Bounded vs Unbounded Tasks.
  3. Manage AI "as a teammate, but not a human one." Management skills transfer to AI; "we're all managing multiple AI agents in our job or will be." Keep deliberate friction so people know they're managing AI and don't cognitively offload consequential decisions — lawyers are still being sanctioned for hallucinated citations years after ChatGPT.
  4. Efficiency-only AI strategy is an opportunity cost. Two questions matter more: what do you do with the gains? and how does AI strengthen the human side of the business? Salesforce's answer to the first: redeploy service people who know the products into forward-deployed engineering (helping customers use AI, not just answering questions). To the second: AI management coaches and nudges ("your survey says your team wants recognition and they just shipped — go acknowledge them"), which converted her from sceptic.
  5. Entry level is being reinvented, not abolished. "It does not make sense long-term for companies not to have talent that is going to be developed into their more senior roles." Salesforce answered the 2026 pullback in graduate hiring by hiring ~1,000 new grads and interns that year; her three interns out-prompted the team on a prompt-injection problem. Managing AI is part of the new entry-level job — the modern "mail room" is using AI to learn the business. See Deskilling Trap (Juniors).
  6. Role redesign runs through HR, with AI mapping tasks to skills. Salesforce has an HR division that sits with each org, uses AI to map tasks to skill sets, sees which tasks are "rising as human tasks", and designs roles of the future (most roles on her team — e.g. responsible AI architect — didn't exist two years ago). Give teams a seat at the table: people closest to the work know where AI fits. Extreme version: Melissa Valentine's Flash Teams — skills made legible, teams assembled and re-formed fast.
  7. A bias toward the future, not the past. Internal AI talent marketplaces work (Seagate, Mastercard: a government-affairs person takes a gig on the security team and lands a role there; hackathons + AI workshops → forward-deployed engineers). The real blocker is cultural: managers hire for pedigree — "no one has 10 years of experience with all of these different skills". Assess what people can create.
  8. Pick a few big bets on role transformation, beyond "give everyone a token budget" ("we all blew through that budget"). Salesforce: engineering is "completely transformed", so every function supporting engineering (including trust/ethics) must re-tool too. Pharma would differ (AI-for-science on top of horizontal use cases).
  9. AI slop is answered with accountability + knowing where not to use AI. "Your work product as an individual is your work product" — stand behind every word regardless of AI. Innovation example: IKEA's team kept getting boxy couches because AI "reverts to the mean" without direction; they front-loaded the brief (human brainstorming on "campfire", "gathering space") before co-ideating with AI → a 10-lb "couch in a box" exhibited in Copenhagen.
  10. Sales and marketing rebalance. AI works the thousands of inbound leads no rep could reach and personalises outreach; complex B2B deals (misaligned stakeholders, reorgs, political pressure) stay relational — customers won't reveal context without trust. Marketers now must make messages legible to the agents customers deploy (see Agent-Readable Artifacts) and reinvent to stand out. Claimed stat: salespeople experience depression at 3× the rate of other professionals (unsourced in the episode).
  11. HR is the "linchpin" of AI transformation — it's a people transformation. Good uses: manager nudges, evidence-based performance management "as opposed to my most recent impression of my employee", skills-to-opportunity matching. Risk: well-known AI-screening failures that reject qualified candidates — "it requires intentionality".

Assessment

  • Strength: concrete, operator-level mechanisms (HR task-mapping division, 1,000-grad counter-cyclical hire, service → FDE redeployment) rather than abstractions; directly corroborates the Economist's framing of Salesforce as the internal-mobility exemplar.
  • Weakness / independence: Goldman is a Salesforce executive promoting a book; Salesforce sells the AI service agents and HR tooling she describes. Every example is either Salesforce or reported second-hand (Seagate, Mastercard, IKEA). No numbers on outcomes (placement rates, retention, service quality). The ~70% figure is a rhetorical heuristic, not a measured split. See the skin-in-the-game caveat on Salesforce.
  • On jobs: "So far, I don't think [fewer people] has been the case, but … we really have to prepare for disruption" — hedged, consistent with AI Adoption and Headcount Growth.

Why it matters for an enterprise IT talent program

  • The delegation question is now a design deliverable. "Just get people licences, give them a token budget" was phase one; phase two is deciding, function by function, what's AI-run, joint, or human-reserved — and writing that down. Pairs with Manager as Translator.
  • The entry-level answer is a hiring decision, not a lament. Salesforce's 1,000-grad hire + "managing AI is the new mail room" is a concrete, citeable counter-position to the Deskilling Trap (Juniors) fear, and matches AI Adoption and Headcount Growth.
  • Pedigree bias is the hidden blocker to internal mobility. Any reskilling / talent-marketplace program fails if hiring managers still screen for "10 years of X". Assess demonstrated artefacts instead — see Bias Toward the Future (Hiring).
  • New roles are real and named (responsible AI architect, forward-deployed engineer) — useful evidence for job-family redesign and for Tasks to Responsibilities Shift.

Connections

Sources

  • Raw: source — HBR IdeaCast bonus episode, transcript via hbr.org, clipped 2026-09-27