AI made the deliverable cheap, and the judgment expensive.
You look at a polished strategy deck that just landed on your desk. Forty pages. Clean structure. The right citations. The right shape. A year ago, that would have told you something.
Now you look at it, and there’s something off you can’t quite name. The deck is good, but it’s also something you could have produced yourself in an afternoon with Claude or ChatGPT. Not perfect, and maybe not in a form you’d want to put your name on.
Polish used to be evidence that hard thinking happened. That disconnect you’re feeling is the realization that it isn’t anymore.
The artifact used to hide the labor
For most of knowledge work history, the deliverable and the thinking were intertwined. A two-week deck meant two weeks of work. A month-long report meant a month of research. The work product was the receipt that proved the labor.
That intertwining is why org charts looked the way they did. Junior analysts produced research. Mid-level managers produced summaries. Senior leaders produced recommendations. Each layer added a markup for the asset, because the asset was hard.
It is also why consulting firms could charge what they charged. Why MSPs priced by ticket volume. Why federal contractors priced by the hour. Why agencies priced by deliverable. The asset was the unit of value, because the asset was the unit of effort.
AI didn’t kill knowledge work; it just made the hidden parts visible. The hardest part of knowledge work used to be knowing what to type.
The value moved
AI has made us more productive, and more importantly, it has made deliverables nearly free.
The cost curve on production has collapsed to near zero. The cost curve on judgment and taste, if anything, became more expensive.
A friend confided in me that of the four proposals he received for a recent RFP, three had structurally identical executive summaries. Same headings. Same hedged confidence. Same competence floor. The differentiation that used to live in the deliverable had migrated somewhere else, and the procurement system had not caught up.
That pattern is showing up everywhere. Five examples, across functions, that anyone running an organization right now will recognize:
- A proposal can be drafted faster than a project manager can review it.
- A policy can be generated faster than legal can validate it.
- A prototype can be created faster than the business can decide whether it should even exist.
- A system assessment can be generated in an afternoon that used to take three weeks, and the bottleneck is now aligning on who owns the fix.
- A board deck can be assembled faster than the leadership team can agree on what it should recommend.
In every case, the bottleneck moved – upstream to framing the problem. Downstream to validating the answer, owning the decision, and making it stick in an organization that might be resistant to change.
The artifact got cheap. The judgment got expensive. And, the pricing has not caught up.
The research backs the shift
A recent study from Christoph Riedl at Northeastern and Ben Weidmann at University College London looked at human-AI collaboration using benchmark data from 667 participants. The testing consisted of math problems, physics problems, and moral reasoning problems. At first, participants had to work solo, and then with an AI assistant.
The researchers wanted to know what ability predicted who would do well with AI.
What they found was that raw problem-solving ability wasn’t the predictor. The predictor was the ability to take perspective, give context, and anticipate what information was missing.
The researchers call it Theory of Mind. Everyone else calls it, “being a good collaborator.”
AI does not simply reward intelligence. It rewards collaboration.
This is why the output stopped being the moat. The output is what the AI helps produce. The moat is the human ability to frame the problem, supply missing context, challenge the answer, and decide what deserves to survive.
The skill that matters most in the AI era is the one we have been undervaluing the longest.
The Four Pressures
When production collapses, four things break. Each one looks operational. Each one is really about value.
1. Hiring: Stop screening for output volume.
The new high-leverage hire is not the person who can produce the most. It is the person who can frame the right problem, give the right context, and tell you when the AI’s answer is wrong. That is not a junior skill. It is not even a technical skill. It is a judgment skill, and most job descriptions still are not testing for it.
2. Pricing: Stop selling production as the scarce part.
You can absolutely still sell deliverables. You just cannot price them as if generating the deliverable is what the client is paying for. Federal procurement still asks for level-of-effort and hours by labor category. MSP contracts are still priced by seat or ticket. Agency retainers still bill by output. All three are asset-economy pricing structures attached to a post-asset world. The point here is that these models will face pressure as buyers realize the visible deliverable is no longer the hard part.
3. Measurement: Stop counting outputs, start counting decisions.
You can no longer measure AI rollouts by volume generated. The new scorecard is whether the work changed a decision, reduced risk, eliminated rework, or solved a real operational problem. Output volume is the old metric of an asset economy.
4. Training: Stop teaching tools, start teaching judgment.
Most AI training programs are tool training in disguise. They teach people to generate, which is the easy part. The hard part, and the part the research actually points out, is teaching people to frame problems and review outputs with judgment. Prompt training is fine. Tool training is fine. But neither one teaches the skill that Riedl and Weidmann found predicted success.
Every one of these is a value problem disguised as an operating problem.
What leaders should do, starting tomorrow
Three important moves:
- Audit what you are paying for. Walk through your last ten major deliverables, internal or external. For each one, ask whether the value was in producing it, or in the thinking that decided it should exist. If you cannot answer, you have found a budget line that is about to come under pressure.
- Audit how you are measuring your team. If your dashboards count outputs, you are measuring the wrong thing. Replace output metrics with decision-quality metrics: decisions made, decisions reversed, and decisions defended successfully under scrutiny.
- Audit how you are training your people. Most AI training programs teach tools. The actual skill that predicts AI success is collaboration. This comes in the form of taking perspective, giving context, and anticipating gaps. That’s a significant shift in curriculum, and most organizations have not written it yet.
The leaders who are successful in the next three years may not have the best AI tools. The tools commoditize fast. They will be the ones who rebuilt their organizations around the work that did not get cheaper.
AI made the artifact cheap and fast. And because it is so fast, judgment has become expensive. The organizations still paying for production are about to find out what they were buying all along.
The work is getting honest
Production has gotten cheap, but AI did not make knowledge work disappear. It made what was always true visible. Judgement and the ability to collaborate with AI systems to produce stronger outputs are now your most valuable skillsets.
The deliverable is not the work anymore. It probably never was.
If you’re ready to rethink the way you work, Doyon Technology Group can help. We provide AI strategy roadmaps and training services to help you on your way to a more productive organization. Connect with us today to request an intro call with our team, and let’s see where this can go.
––––––
About the Author
Greg Starling serves as the Head of Innovation & Growth for Doyon Technology Group. He has been a thought leader for the past twenty years, focusing on technology trends, and has contributed to published articles in Forbes, Wired, Inc., Mashable, and Entrepreneur magazines. He holds multiple patents and has been twice named as Innovator of the Year by the Journal Record. Greg also runs one of the largest AI information communities worldwide.
Doyon Technology Group (DTG), a subsidiary of Doyon, Limited, was established in 2023 in Anchorage, Alaska to manage the Doyon portfolio of technology companies: Arctic Information Technology (Arctic IT®), Arctic IT Government Solutions, and designDATA. DTG companies offer a variety of technology services including managed services, cybersecurity, and professional software implementations and support for cloud business applications.