Refacto

Podcast episode

S2E13: Thinking about Consulting? | Round table with James Deaker aka The Yield Doctor

by Ken Rona · ·Leadership In

ai-in-adtech career-development consulting yield-optimization

TL;DR

A career-focused roundtable on independent consulting between hosts Geoff Wolinetz and Greg MacDonald and guest James Deaker (founder, Kurokea Media; "The Yield Doctor"), all active ad-tech consultants. The episode is primarily professional-development content — pros/cons of going independent, AI's encroachment on consulting work, and how to structure client engagements — with light but real relevance to ad-tech operators thinking about talent and vendor strategy.

What was covered

  • Defining the consulting landscape: James Deaker, who holds a Stanford PhD and held pricing/yield leadership roles at Yahoo, Verizon, and Microsoft, framed three flavors of independent work — traditional consulting (expertise-driven, ongoing), advisory (opinion/experience, often project-based), and fractional (embedded part-time leadership, sometimes with direct reports). He noted the post-pandemic shift to fully remote engagements has materially changed the model.
  • Deaker's Lyft engagement as a case study: He described an active ~one-day-a-week pricing and yield advisory role with Lyft's advertising business, which started as a defined 10-week project and organically expanded — illustrating both the fractional model and scope-creep dynamics.
  • AI as a double-edged force for consultants: All three discussed AI (tools like Claude cited by name) as simultaneously making consultants more productive and giving clients false confidence they can solve complex strategy problems without outside expertise. Deaker described a client who fed a pricing problem to an AI agent, got an answer, and nearly bypassed the more important step of defining the right objective function.
  • The "AI slop" risk in strategy work: Geoff Wolinetz raised the point that current AI models homogenize outputs — pulling from prior work — making them poorly suited to genuinely novel ad-product innovation. Deaker agreed and noted that "vanilla consulting" with low differentiation will likely be displaced.
  • Building vs. staying solo: Deaker argued that scaling to a team requires repeatability (cookie-cutter projects), which limits the type of work a specialist can take on. He cited a 2007 Facebook pricing project (when Facebook was small) as a formative lesson: clients buy the named expert, not the junior team.
  • Business development realities: Wolinetz noted that 80–90% of consulting work is unpredictable BD, not billable hours, and that the market reads ambiguity about your intent (consulting vs. job-seeking) as a reason not to send referrals.
  • Resources cited: David A. Fields Consulting (newsletter and book), Bradley Jacobs / Mylance, and Fractionals United.

Notable claims & predictions

  • James Deaker: "I would say 80 to 90% of my consulting work over the last three years has come through either my YouTube channel or the LinkedIn promotions that I do of the videos." — Points to content-led BD as a viable and underused pipeline for independent ad-tech consultants.
  • James Deaker: "The vanilla consulting that a lot of companies have managed to subsist on is going to go away. The stuff where there actually isn't that much value add." — Predicts AI displacement of undifferentiated consulting, accelerating the flight to specialists.
  • Geoff Wolinetz (paraphrasing a large consultancy's practice): A major consulting firm has begun billing AI agents at the same hourly rates previously applied to human consultants — framing machine time as a direct substitute line item.
  • James Deaker on token-based AI pricing: "I caution anyone linking their long-term viability to a token-based model. This is like the telecom models from early 2000s where they charged per minute… I think we're going to see a migration either to subscription-based models or potentially to value-add models." — Forecasts that AI cost structures will shift away from consumption pricing.
  • Greg MacDonald: "AI is effectively designed to homogenize everything… You won't end up with an innovative solution if you just throw it at an AI agent right now." — Argues AI currently cannot replace human expertise in novel ad-product design.
  • David A. Fields (quoted by Deaker): "The most reliable predictor of small consulting firms' growth are: one, the narrowness of the depth of their expertise; and two, the breadth of their rainmaking team."

Fact check

  • Deaker's claim that Facebook was "tiny" in 2007 when he ran a pricing engagement with them: Directionally accurate — Facebook had roughly 50–100 million users by 2007 but was not yet a scaled global ad platform. "Tiny" is relative but not misleading in the context of its current scale. Unverified but plausible.
  • Wolinetz's claim that a large consultancy is billing AI agents at the same rates as human consultants: Presented as hearsay ("I actually heard of…"). Not attributed to a named firm. This is unverified and worth discounting — it may be accurate directionally, but the framing invites listeners to accept it as established industry practice. Treat as anecdote, not established fact.
  • Deaker's AI pricing prediction (token → subscription migration): Framed as a personal forecast, not a fact claim. The analogy to early 2000s telecom per-minute billing is illustrative but imperfect — AI pricing dynamics differ from bandwidth constraints. Opinion/forecast; incentive to note: Deaker has reason to argue AI won't commoditize human expertise, since that's his livelihood.
  • Deaker's "80–90% of work from YouTube/LinkedIn" claim: Self-reported, unverifiable, but specific enough to be useful as anecdotal evidence for content-led BD. Unverified.

No claims rise to the level of clearly false or demonstrably misleading.

Why this matters for ad-tech operators

  • Talent and vendor strategy: As AI trims headcount and budget for generalist work, the episode reinforces a pattern already visible in ad-tech — buyers of consulting services (publishers, DSPs, SSPs) will increasingly concentrate spend on narrow specialists (e.g., yield optimization, identity strategy) rather than broad advisory firms. Operators evaluating consultants should probe whether the expertise they are paying for is genuinely defensible against AI substitution.
  • AI agent risk in pricing/yield workflows: Deaker's anecdote about a client who used an AI agent to "solve" a pricing problem without defining the objective function is a cautionary note for ad-tech operators running yield optimization or rate-card projects internally. Badly framed AI outputs in this domain can lock in infrastructure that underperforms for years.
  • Lyft advertising as a scaling example: Deaker's active engagement with Lyft's ad business signals that ride-share and mobility platforms continue to build out programmatic ad monetization, creating demand for external pricing and yield expertise — a market-structure data point for SSPs and DSPs tracking emerging inventory sources.
  • Impact is indirect: This episode is primarily career/professional-development content. Ad-tech executives and operators should treat it as a talent-market and vendor-selection lens, not a source of platform, M&A, or regulatory intelligence.

Analysis

Showing the shorter version.

Thinking About Consulting? What the Yield Doctor Actually Said

A career roundtable with Geoff Wolinetz, Greg MacDonald, and James Deaker (founder of Kurokea Media, former Yahoo/Verizon/Microsoft pricing chief, self-styled "The Yield Doctor") ended up being less about career advice and more about what AI does to expensive human expertise in yield and pricing. Operators should pay attention to the second topic.

The demand signal

Deaker's most interesting disclosure: he's been advising Lyft's ad business one day a week for nearly a year. Mobility platforms are building programmatic monetization without the in-house yield muscle to support it. The fastest-growing ad businesses right now aren't traditional publishers; they're apps that happen to have screens and users. That's a live demand signal for SSPs and yield vendors looking for new inventory sources. And it predicts a pattern: these platforms can build the ad server faster than they can hire the pricing brain, so they reach for fractional specialists instead.

What AI actually kills

Deaker's thesis: AI wipes out "vanilla consulting," the generalist advisory layer, and anything that's really just analysis in disguise. The expensive named specialist survives because clients are paying for judgment, and a model doesn't have that. The junior team doing repeatable analysis is what gets replaced, cheaper and faster by a model. That compresses the middle of the consulting market, and for operators, it means the smart spend is fewer bodies and more targeted specialist days on decisions that lock in revenue for years.

Take the incentive into account before accepting this fully. Deaker is a yield consultant arguing AI can't replace yield consultants. He's not a neutral witness.

The risk operators are importing right now

The most useful thing Deaker said had nothing to do with careers. He described a client who fed a pricing problem to an AI agent, got a clean answer, and nearly shipped it without ever defining the objective. Price realization? Margin? Package clarity? Different objectives, different answers, and the model happily optimized for whichever one it guessed.

In yield work, that's not a bad quarter. That's infrastructure. You set floors, tiers, and packaging around a wrong objective and you're living with it for a year or two before the underperformance is even visible. The AI didn't fail. The framing did. That's the gap operators are skipping over when they run rate-card and yield projects on Claude in-house and skip the expensive human who asks "what are we actually solving?"

The operating rule this points to: before you ship any AI-driven yield or rate-card output, make sure a person defined the objective function before the model ran. If nobody can name it, the number is decoration.

One cost-planning note

Deaker flags token-based AI pricing as the early-2000s per-minute billing model: it probably migrates to subscription, and your unit economics move with it. If you're building workflows on consumption pricing, plan for that shift.

Our call

By the 2027 upfront season, at least one more non-traditional platform in mobility, delivery, or ride-share beyond Lyft will publicly stand up or materially expand a programmatic ad business and lean on outside pricing/yield expertise rather than build it in-house first. Confidence: medium. The demand pattern is real; which platform and when is harder to pin. We're wrong if no platform beyond Lyft surfaces that pattern, or the ones that scale do it with in-house yield teams from day one.

Listen to the episode →

Also covered this issue

Comments