Ask a room of partnership leaders what they are doing about AI and you will hear some version of the same answer: we are automating. Partner comms drafts, QBR decks, deal reg reviews, account mapping. Someone mentions an agent pilot. Someone mentions prompt training. All of it is real work, and most of it is worth doing.
The question every partner team is asking is how do we automate. The question very few teams are asking is what part of our business model stopped working. Those are different questions, and the gap between them is where partner programs are going to run into trouble over the next two years.
The economics moved before the task list did
Here is the pattern worth naming. AI shows up inside your workflow, so you treat it as a workflow problem. You audit tasks, automate the tedious ones, and bank the saved hours. Leadership sees a productivity story. It feels like progress, and at the task level it is.
Meanwhile, the money is moving somewhere else. We coded every major partner program announcement from hyperscalers and large platform vendors over the past 12 months, 20 in total, by a single question: where did money, credit, or deal flow move?
A few examples from that dataset show what this looks like in practice.
Microsoft is changing what counts as a deal. Azure IP co-sell is moving to a Marketplace-first model, where verified marketplace transactions replace self-reported co-sell. At the same time, an AI discovery interface lets customers find software with natural-language prompts. Read that from the attribution side: the thing that sources the deal is increasingly a search layer the vendor owns. That has direct consequences for deal registration logic and for any sourced-revenue number built on it.
AWS and Dell are changing what gets paid. AWS's Business Value Realization motion ties funding to customer-attested milestones, and Specialization renewals now require launched opportunities rather than certifications alone. Dell's 2026 program updates pay on AI outcomes and transactions. The pattern across both: incentive dollars are being retagged from activity to outcome.
Anthropic is changing what a tier is worth. The Claude Partner Network launched open and free, and drew more than 40,000 applicants with over 10,000 consultants certified within months. When certification is free and instant, badge counts stop being a scarce asset. Tier structures built on headcount and certifications lose their pricing power.
None of this was announced as a program teardown. It arrived as product news and incentive updates. But taken together, the revenue logic that assumed a human sourced, registered, and closed the deal is being rewritten one announcement at a time.
Why the automation question feels safer
How do we automate is an answerable question. You can run the pilot, count the saved hours, and show the slide. Early signals from our State of Partnerships survey point at something sharper underneath: a meaningful share of partnership professionals report using AI more than their leadership knows. The automation is already happening. What is missing is a strategy for the economics.
What part of our business model stopped working is a harder question because the answers implicate the structures you run: your tiers, your deal reg, your comp plan, and the sourced-revenue number your CFO already questions.
That raises the question this series will keep pulling on. If the model that assumed a human sourced the deal is going away, whose revenue is it now? When an agent surfaces the opportunity, one partner influences it, another embeds in the workflow, and a third delivers the outcome, who gets paid? Very few teams have a defensible answer today. Weeks 10 through 12 of this series are dedicated to building one, and by the end of the season I will introduce what I believe is the missing accounting layer for ecosystems. I call it the Outcome Ledger. More on that as we go.
Monday Morning Move: run the audit that actually matters
Twenty minutes. List your team's ten most time-consuming recurring tasks and sort each into one of three buckets.
| Bucket | Test | Operating implication |
|---|---|---|
| Automate | An agent can do this end to end today with acceptable risk | Stop staffing it. Redeploy the hours this quarter. |
| Augment | AI drafts, a human judges | Your quality bar becomes the differentiator. |
| Defend | The value is the human: trust, negotiation, judgment, authority | This is your job description in 24 months. Invest here. |
Then the step most teams skip. For every task in the Automate column, write down the revenue logic attached to it. Deal reg reviews carry sourced-revenue credit. QBR prep carries the partner health narrative. Comms carry engagement metrics. When a task is automated, the number attached to it does not disappear. It becomes unowned, and unowned numbers are where attribution fights start.
Keep the list of unowned numbers. It may be the most useful document your team produces this quarter, and we will come back to it in Week 10.
One more thing about that Defend column. The skills that live there are exactly the ones hiring managers are starting to pay a premium for, and demand for AI-fluent partner operators is already outrunning supply. We are analyzing 300+ partnership job descriptions to show precisely which skills are moving up and which are moving out. That is Week 5.
Download the task audit templateEvery episode of this series makes one dated, falsifiable prediction, and we will score all of them publicly at the end of each year. Here is the first: by the end of 2028, outcome-attested funding will be a standard component of every major hyperscaler partner program, and self-reported co-sell will no longer earn incentive dollars at any of them. Microsoft's Marketplace-first shift and AWS's BVR motion are the first two data points. Hold us to it.
- Microsoft, Marketplace-first co-sell. Azure IP co-sell moves to the Marketplace. The marketplace is becoming the system of record for what counts as a partner deal. Co-sell motions that live in spreadsheets and Slack threads will struggle to show up in that record.
- AWS, agents in the buying motion. Marketplace Agent Mode brings conversational procurement, and the consulting listing fee drops from 2.5% to 0.5%. AWS is pricing services deal flow toward the marketplace while putting an agent at the front of the buying experience.
- Distribution concedes the transaction layer. Distributors now describe hyperscaler marketplaces as central to partner growth. The open question is who owns the customer data once the transaction moves.
Score your team honestly. Each episode adds a dimension, and in Week 13 the full maturity model goes public.
- Level 1. Individuals are experimenting with AI tools. No team strategy.
- Level 2. We are automating tasks and tracking saved hours.
- Level 3. We have mapped which parts of our revenue logic AI changes.
- Level 4. We are redesigning tiers, comp, and attribution for the new economics.
Most teams we talk to sit at Level 2 and describe themselves as Level 4.
Google put $750 million behind its partners' agentic AI development. AWS put agents inside Partner Central. Both read like product news. They are better understood as the funding model of the old channel being rebuilt in public. Episode 2: The $750M Signal.
One ask. We are building an operator-led benchmark on how partnership teams actually use AI. Seven minutes, anonymous, results published publicly. Add your voice.