Latest / Uphoff on Media Podcast / Your Agentic AI Isn’t Failing. Your Math Is

Transcript

Your Agentic AI Isn’t Failing. Your Math Is · Uphoff on Media Podcast

• 0:01
  SPEAKER_00: Hey, it's Tony.
• 0:02
  SPEAKER_00: It's fascinating to see how fast the use of agentic AI in B2B businesses has mainstreamed.
• 0:09
  SPEAKER_00: Having had a ringside seat at every technology disruption of the last 30 years,
• 0:13
  SPEAKER_00: I can tell you that I've never seen anything remotely close to the speed of adoption
• 0:18
  SPEAKER_00: and implementation of agentic AI.
• 0:21
  SPEAKER_00: Part of this is the ease of trial ability and experimentation.
• 0:26
  SPEAKER_00: Previous technology ways required a significant investment in the technology,
• 0:29
  SPEAKER_00: people, and services before you could get anywhere close to implementation.
• 0:34
  SPEAKER_00: This ease of use also helps to deliver impact faster as individual users
• 0:39
  SPEAKER_00: or job functions can identify and build a use case without needing to go through
• 0:43
  SPEAKER_00: an enterprise-wide rollout and adoption curve.
• 0:47
  SPEAKER_00: The challenge I'm seeing now is that businesses are struggling to standardize
• 0:51
  SPEAKER_00: a measurable ROI formula for agentic AI.
• 0:54
  SPEAKER_00: As a result,
• 0:55
  SPEAKER_00: some are struggling to set expectations at the board,
• 0:58
  SPEAKER_00: corporate, and individual level for AI use and investment.
• 1:02
  SPEAKER_00: There's a bit of history repeating itself here.
• 1:04
  SPEAKER_00: As with other technology innovations,
• 1:06
  SPEAKER_00: many business leaders are looking at the technology itself in determining the ROI.
• 1:11
  SPEAKER_00: Measuring technology is a fool's mission.
• 1:14
  SPEAKER_00: The true test of technology is the impact of the capability it gives you.
• 1:18
  SPEAKER_00: Nothing else matters.
• 1:20
  SPEAKER_00: Not processing speed, not MIPS, not queries per second.
• 1:23
  SPEAKER_00: It is demonstrable impact on your business capabilities.
• 1:27
  SPEAKER_00: That's what this episode is about.
• 1:29
  SPEAKER_00: How do you measure ROI on agentic AI and also how do you set expectations
• 1:34
  SPEAKER_00: for each stage of adoption?
• 1:37
  SPEAKER_00: I'd encourage you to visit Uphoff on Media and Substack and view this full post
• 1:42
  SPEAKER_00: as well.
• 1:42
  SPEAKER_00: In it,
• 1:43
  SPEAKER_00: you'll find a downloadable framework called ADAC where you can use in your business
• 1:49
  SPEAKER_00: to manage agentic AI adoption and ROI.
• 1:52
  SPEAKER_00: Okay.
• 1:53
  SPEAKER_00: Let's get to work.
• 1:55
  SPEAKER_00: This episode is titled, Your Agentic AI Isn't Failing, Your Math Is.
• 2:00
  SPEAKER_00: Think big, start small, and never fund the next stage without calculating the payback.
• 2:07
  SPEAKER_00: MIT's NANDA initiative found in 2025 that 95%
• 2:12
  SPEAKER_00: of enterprise generative AI pilots delivered no measurable P&L return against
• 2:18
  SPEAKER_00: an estimated $30 to $40 billion invested.
• 2:21
  SPEAKER_00: Gardner's forecast for what happens next is worse.
• 2:25
  SPEAKER_00: More than 40% of agentic AI projects will be canceled by the end of 2027,
• 2:30
  SPEAKER_00: driven in part by escalating costs and unclear business value.
• 2:36
  SPEAKER_00: The pilots that don't quietly die become the projects that get killed later.
• 2:40
  SPEAKER_00: It'd be easy to mistake this data as proof that technology isn't ready.
• 2:44
  SPEAKER_00: It's not the technology, it's your business case.
• 2:48
  SPEAKER_00: I have managed through enough disruptive transformations to recognize this pattern immediately.
• 2:53
  SPEAKER_00: I watched leaders obsess over MIPS when computing capacity was the scarce resource.
• 2:59
  SPEAKER_00: I watched them obsess over uptime and query volume during the cloud transition.
• 3:03
  SPEAKER_00: Now it's tokens processed and API calls.
• 3:07
  SPEAKER_00: The instrument changes, the mistake doesn't.
• 3:09
  SPEAKER_00: Leaders measure the technology.
• 3:11
  SPEAKER_00: They rarely measure the capability it creates.
• 3:14
  SPEAKER_00: Agentic AI didn't create that dynamic, it just made it faster and more expensive.
• 3:20
  SPEAKER_00: This isn't a tool you adopt, it's how you transform your business.
• 3:25
  SPEAKER_00: I've led through two of these technology disruptions as an operator,
• 3:28
  SPEAKER_00: print to digital and digital to platform,
• 3:30
  SPEAKER_00: and I'm managing through the third right now.
• 3:33
  SPEAKER_00: Platform to machine, agentic systems that don't just deliver information, they act on it.
• 3:38
  SPEAKER_00: The lesson repeats every cycle.
• 3:40
  SPEAKER_00: The shift looks survivable to the leaders defending the old form and inevitable to
• 3:45
  SPEAKER_00: the ones who see the function underneath and build for it early.
• 3:49
  SPEAKER_00: The leaders who move first aren't smarter.
• 3:51
  SPEAKER_00: They just stop asking whether the shift is real and start asking what to build on it.
• 3:56
  SPEAKER_00: This one is not a fad and it's not optional.
• 3:59
  SPEAKER_00: It will touch every function in business from sales to finance to operations to product.
• 4:04
  SPEAKER_00: To be clear, because it's where most companies lose the thread.
• 4:09
  SPEAKER_00: Agentic AI itself was never the point and the technology wasn't in the three shifts
• 4:14
  SPEAKER_00: before it either.
• 4:15
  SPEAKER_00: The point is what you build with it and how differently you run the business because it exists.
• 4:21
  SPEAKER_00: Confuse the tool with the outcome and you'll spend two years in
• 4:24
  SPEAKER_00: a real budget proving nothing at all.
• 4:28
  SPEAKER_00: Efficiency is where good intentions go to die.
• 4:31
  SPEAKER_00: Most agentic AI initiatives get funded on a word that sounds like a standard
• 4:35
  SPEAKER_00: and really isn't, efficiency.
• 4:38
  SPEAKER_00: Efficiency is whatever the person reporting it needs it to be that quarter.
• 4:42
  SPEAKER_00: Hours saved,
• 4:43
  SPEAKER_00: tickets closed,
• 4:44
  SPEAKER_00: adoption rate,
• 4:45
  SPEAKER_00: none of it is wrong to track and none of it survives a P&L review,
• 4:49
  SPEAKER_00: which is exactly the gap Gartner is describing when it names unclear business values
• 4:54
  SPEAKER_00: the top reason these projects get killed.
• 4:57
  SPEAKER_00: The only measurement that matters is whether a technology improves your business, period.
• 5:04
  SPEAKER_00: Here's what that looks like in practice.
• 5:06
  SPEAKER_00: Agentic AI success is defined by direct commercial impact,
• 5:11
  SPEAKER_00: a measurable, attributable improvement in revenue, margin, or cash conversion.
• 5:17
  SPEAKER_00: Not adjacent to those numbers, not a leading indicator of those numbers someday.
• 5:22
  SPEAKER_00: Those numbers named in advance with a baseline isolated from everything else moving
• 5:26
  SPEAKER_00: in the business at the same time.
• 5:29
  SPEAKER_00: I've learned this discipline decades before large language models existed.
• 5:33
  SPEAKER_00: Pipeline 360 EBITDA margin didn't move because we picked a new technology.
• 5:37
  SPEAKER_00: It moved because we named the number before we started,
• 5:40
  SPEAKER_00: built a plan around hitting it,
• 5:41
  SPEAKER_00: and checked it on a calendar we'd committed to in advance.
• 5:44
  SPEAKER_00: Eugenic AI doesn't change that discipline.
• 5:47
  SPEAKER_00: It just raises the cost and risks of skipping it.
• 5:52
  SPEAKER_00: Think big, start small, measure what you can actually prove.
• 5:56
  SPEAKER_00: Two instructions in this order, and most companies only manage one of them.
• 6:00
  SPEAKER_00: Think big.
• 6:01
  SPEAKER_00: The ambition has to match what's actually happening.
• 6:04
  SPEAKER_00: Agentic AI will change how your business runs and grows,
• 6:07
  SPEAKER_00: not which vendor logo sits in your tech stack.
• 6:11
  SPEAKER_00: If your plan tops out at a chatbot on the website,
• 6:14
  SPEAKER_00: you're thinking too small to justify the investment.
• 6:17
  SPEAKER_00: Start small.
• 6:18
  SPEAKER_00: One workflow, one named owner, one proof point.
• 6:22
  SPEAKER_00: Define before you spend a dollar.
• 6:24
  SPEAKER_00: The companies pulling ahead right now aren't running the most pilots.
• 6:27
  SPEAKER_00: They're the ones willing to kill nine to get one into production with a number attached to it.
• 6:33
  SPEAKER_00: The gap between those two instructions,
• 6:36
  SPEAKER_00: big ambition and narrow execution is where most companies fall apart.
• 6:40
  SPEAKER_00: They either think too small,
• 6:41
  SPEAKER_00: a tool instead of a transformation,
• 6:44
  SPEAKER_00: or start too big,
• 6:45
  SPEAKER_00: a company ride rolled out with no single owner and no defined proof point.
• 6:49
  SPEAKER_00: What's missing is a way to know project by project exactly where you stand
• 6:54
  SPEAKER_00: and what has to be true before the next dollar goes in.
• 6:58
  SPEAKER_00: The ladder, assisted, delegated, accountable, compounding.
• 7:02
  SPEAKER_00: I built a framework for exactly this with clients, a four stage ladder called ADAC.
• 7:09
  SPEAKER_00: It doesn't score your company, it scores the initiative in front of you.
• 7:12
  SPEAKER_00: And it works the same way whether you're the one funding it,
• 7:15
  SPEAKER_00: running it, or diligencing it before you invest.
• 7:20
  SPEAKER_00: Number one, assisted.
• 7:21
  SPEAKER_00: A human still owns the task and the result.
• 7:24
  SPEAKER_00: This is where almost every company starts and where most never leave.
• 7:28
  SPEAKER_00: Someone uses an AI co-pilot for research,
• 7:31
  SPEAKER_00: drafting,
• 7:32
  SPEAKER_00: summarizing,
• 7:33
  SPEAKER_00: genuinely useful, not yet a workflow, and not yet anyone's job to make it one.
• 7:38
  SPEAKER_00: The sign that you're stuck here, your only evidence is a usage log.
• 7:42
  SPEAKER_00: Quote, unquote, the team uses it daily as a habit report, not a business case.
• 7:46
  SPEAKER_00: Name an owner this quarter and give them one workflow to rebuild, not tend to experiment with.
• 7:51
  SPEAKER_00: Number two, delegated.
• 7:53
  SPEAKER_00: AI owns the task, a human still owns the result.
• 7:56
  SPEAKER_00: This is where most of Gartner's cancellations happen.
• 7:59
  SPEAKER_00: AI executes a defined piece of work inside a workflow,
• 8:02
  SPEAKER_00: drafting the follow-up,
• 8:04
  SPEAKER_00: qualifying the lead,
• 8:05
  SPEAKER_00: extracting the contract terms, while a human reviews everything before it moves.
• 8:09
  SPEAKER_00: Real progress and also where companies get addicted to running pilots instead of finishing one.
• 8:15
  SPEAKER_00: The tell?
• 8:16
  SPEAKER_00: A portfolio of pilots, none with an exit date.
• 8:19
  SPEAKER_00: Freeze new ones.
• 8:21
  SPEAKER_00: Take the pilot closest to done.
• 8:23
  SPEAKER_00: Name its financial proof point and get it into production or kill it.
• 8:29
  SPEAKER_00: Accountable.
• 8:30
  SPEAKER_00: AI owns the task and the result.
• 8:32
  SPEAKER_00: A human owns the exceptions.
• 8:34
  SPEAKER_00: This is where my sentence stops being aspirational and starts being enforced.
• 8:38
  SPEAKER_00: The workflow runs end to end.
• 8:40
  SPEAKER_00: A human steps in only when something breaks.
• 8:43
  SPEAKER_00: And you should expect an attributable line item on the P&L by now.
• 8:47
  SPEAKER_00: If you can't name it, you're not actually here.
• 8:49
  SPEAKER_00: You're still at delegated wearing this stage's name tag.
• 8:53
  SPEAKER_00: The sign you're stuck,
• 8:54
  SPEAKER_00: the workflow performs in every test,
• 8:56
  SPEAKER_00: but leadership still won't remove the human checkpoint.
• 9:00
  SPEAKER_00: That's a governance problem,
• 9:01
  SPEAKER_00: not a technology one,
• 9:02
  SPEAKER_00: because nobody has agreed who's accountable when the agent is wrong on
• 9:05
  SPEAKER_00: a live customer or on something with real dollar amount.
• 9:09
  SPEAKER_00: Name that owner with defined exposure and a plan to reverse it before you remove
• 9:13
  SPEAKER_00: the checkpoint, not after.
• 9:15
  SPEAKER_00: In compounding, AI owns the result and keeps improving how it gets there.
• 9:21
  SPEAKER_00: The system doesn't just run the workflow,
• 9:23
  SPEAKER_00: it improves it without anyone filing a change request.
• 9:26
  SPEAKER_00: This is where the mechanism I described in the post looped engineering stops being
• 9:31
  SPEAKER_00: a coding practice and starts being an operating one.
• 9:34
  SPEAKER_00: Judgment encoded well enough that the system holds your standard
• 9:38
  SPEAKER_00: without you restating it every time.
• 9:40
  SPEAKER_00: It's rare, and it's the only stage where the return compounds instead of holding flat.
• 9:45
  SPEAKER_00: The sign that you're stuck.
• 9:46
  SPEAKER_00: Improvement stops the day your one champion gets pulled onto something else,
• 9:50
  SPEAKER_00: which means you never built the capability.
• 9:52
  SPEAKER_00: You borrowed one person's judgment for a while.
• 9:55
  SPEAKER_00: The fix is institutional.
• 9:57
  SPEAKER_00: The improvement loop needs a standing role, not a hero.
• 10:02
  SPEAKER_00: Score the project, not the company.
• 10:03
  SPEAKER_00: One warning before you use this stages framework.
• 10:07
  SPEAKER_00: Don't score your company, score the initiative.
• 10:10
  SPEAKER_00: A business can be accountable in sales operations and stuck and assisted in finance.
• 10:14
  SPEAKER_00: And that's not a contradiction, that's normal.
• 10:17
  SPEAKER_00: A founder uses this to decide what to fund next.
• 10:20
  SPEAKER_00: An operator uses it to know where the org actually stands, project by project.
• 10:24
  SPEAKER_00: An investor uses it in diligence to tell a real deployment from a demo.
• 10:29
  SPEAKER_00: Number one, what's the metric, revenue, margin, or cash conversion?
• 10:33
  SPEAKER_00: Number two, what's the baseline, where this number stood before the initiative touched it?
• 10:39
  SPEAKER_00: Number three, what's the required delta, the number this has to move to justify the spend?
• 10:44
  SPEAKER_00: And number four, what's the date, when you'll check it?
• 10:47
  SPEAKER_00: If you can't answer all four before you write the check,
• 10:50
  SPEAKER_00: you're not running a workflow, you're funding a hope.
• 10:54
  SPEAKER_00: None of this requires you to become a data scientist and none of it requires you
• 10:57
  SPEAKER_00: to wait until the technology settles down before you move.
• 11:01
  SPEAKER_00: It won't settle down.
• 11:02
  SPEAKER_00: What it requires is the discipline that has separated companies that make money
• 11:06
  SPEAKER_00: from companies that make noise since long before agentic AI existed.
• 11:12
  SPEAKER_00: Name the number before you start,
• 11:13
  SPEAKER_00: know exactly which stage you're standing in and refuse to fund the next one until
• 11:17
  SPEAKER_00: the last one paid for itself.
• 11:19
  SPEAKER_00: Confused,
• 11:20
  SPEAKER_00: frustrated,
• 11:21
  SPEAKER_00: behind, picked, Whichever fits this week, none of it disqualifies you from starting.
• 11:27
  SPEAKER_00: A 95% failure rate isn't evidence you're late.
• 11:30
  SPEAKER_00: It's evidence almost everyone is still measuring the wrong thing.
• 11:33
  SPEAKER_00: Fix your business case first and you move ahead, not catch up.
• 11:37
  SPEAKER_00: Think big, start small and never confuse activity with proof.
• 11:44
  SPEAKER_00: Hey, that's the episode.
• 11:45
  SPEAKER_00: If you found this useful, please do one thing.
• 11:47
  SPEAKER_00: Follow the Uphoff on Media podcast on Apple or Spotify, and please leave a rating.
• 11:52
  SPEAKER_00: It's the single best way to help another operator find this.
• 11:55
  SPEAKER_00: I'm Tony Uphoff, back at it next week.