Dorian Kominek's Playbook for Making AI Partner Matching Actually Work

Automating partner selection doesn't remove human judgment from the process. It relocates it — and most teams are automating before they've worked out where it went.

Ask most partner leaders what AI is doing to go-to-market right now and you'll get a confident answer. Dorian Kominek gives a different one.

"Nobody knows. We're all scouring LinkedIn for what to do next. There's been a lot of promises made, and yet there still is sort of a lack of traction that I think everybody's feeling." — Dorian Kominek, Channel Leader, Zapier

Kominek spent a decade as an entrepreneur before moving into tech and landing in partnerships. They now lead channel at Zapier, which puts them in an unusually recursive position: building an AI-assisted partner motion, for a company whose partners make their living implementing automation for everyone else.

That vantage point produces a more useful playbook than most AI-and-partnerships content, because it starts with what breaks rather than what's possible. What follows are the four things they'd fix in order.

1. Solve the human problem before you build the workflow

The first blocker isn't technical, and Kominek is emphatic that no amount of tooling routes around it.

A co-sell motion needs cooperation from the entire sales org — leadership, managers, and the individual account executives who have to actually run it. Which means the design questions are the seller's questions, not the partner team's: what kind of leads do they want from partners, what workflows have to exist for them to accept a partner-sourced deal, and what about this is going to be genuinely exciting to them.

Partner people skip this step because from where they stand the value is obvious. Partners bring warm leads. Those leads arrive half-sold. The AE closes more, hits quota sooner, earns more commission. It's the same story in reverse when you pass deals out.

But an AE's day has no capacity for obvious.

"Adding any extra layer or something that could be perceived as friction to their day-to-day is gonna be a no-go. They're not even gonna listen to you." — Dorian Kominek, Channel Leader, Zapier

Which is why leadership buy-in is the actual prerequisite. Not as a budget approval, but as evangelism — so the motion is being advocated for from above rather than sold one AE at a time. The best co-selling programs Kominek has seen all shared it: leadership that understood partnerships, and excitement rather than trepidation. Once that exists, they say, you can operationalise anything.

They're equally clear about the structural version of the same problem. A channel motion that sits outside the revenue org, or reports somewhere underneath it, will struggle regardless of buy-in — because channel works by riffing off the sales org's pre-built go-to-market engine. If you can't see that engine, you can't replicate it.

2. Understand where the human-in-the-loop step actually went

This is the core insight, and it's the one most teams miss.

Zapier's channel team is building a co-sell lead-passing workflow where outbound deals are matched to partners with near white-glove precision: the tools and tech stack a partner works in, the languages they speak, their geography, what they say about themselves in their bio and on their site. An algorithm works out best fit.

The upside is straightforward — better customer outcomes, higher likelihood to close, and far less manual correction. But the part Kominek keeps returning to is what it removes: the requirement that partner managers hold tacit knowledge of the entire partner directory in their heads. That knowledge never scaled, and it walked out the door whenever a partner manager did.

The catch is that the tacit knowledge was doing real work. It was quietly absorbing bad partner data — a human looking at a shortlist and saying no, that one.

Their illustration is the best in the recording. Imagine looking for a psychologist on a directory site where every practitioner claims to specialise in every condition. The information isn't quite false; it's just unusable, because it can't all be true and therefore none of it helps you choose. That is what most partner directories look like.

With a human reviewer at the end, that's survivable. With an agent making the selection, it isn't.

"That's where the human-in-the-loop step moves upstream to data validation." — Dorian Kominek, Channel Leader, Zapier

So the job doesn't get eliminated. It changes shape — from routing deals to guaranteeing that what flows into the model is true.

3. Let customers decide which data you validate

Once validation is the job, the question becomes which fields are worth the effort. Kominek's answer is refreshingly external: whatever your customers are asking for.

At Zapier, that meant tech stack, and specifically evidence of it. Which tools do you work with, and can you show us examples of what you've built with them? Not a self-reported capability list — proof.

They separate the fields by how hard they are to keep honest. Languages and geography are easy to control. Tech stack, tooling, actual demonstrated skill, and vertical specialisation — RevOps versus People Ops versus Marketing Ops — take genuine maintenance, because that's where the temptation to over-claim lives.

Which makes partner behaviour part of the data strategy. Partners have to be actively managed away from claiming they do everything and work with every app on the planet.

"They can't just, sort of, hand-wavy claim that they are anymore." — Dorian Kominek, Channel Leader, Zapier

And the same logic scales down. If your customers need Marketing Ops depth, that's the field you validate. The customer's question defines the schema.

4. Measure adoption before you measure revenue

Kominek has taken a deliberately iterative approach — small steps rather than a single launch — and the metrics they watch change by phase.

Early on, none of them are revenue. They're adoption: how many account executives have used the workflow, how many unique individuals, how many times each, how many deals have moved through it. And the diagnostic signal underneath — are team members routing around it, finding some other path of least resistance? That's not a revenue problem. That's an enablement gap.

The reasoning is blunt and correct. You can build something genuinely capable of driving revenue, and if nobody wants to use it, you will never find out.

After adoption, their leading indicator is customer proof. It's very hard to argue with a customer saying the partner did the most for them, and the attached seller becomes an advocate for free. They note the inverse arrives faster — if it went badly for a customer, you hear about it immediately, which makes it an unusually honest early signal.

Only then do the familiar comparisons matter: partner-attached closed-won versus deals without a partner attached, and further out, churn and stickiness. In their experience those all move the same direction. Churn lower, close rate higher, cycle time shorter.

One more thing: you may not need the agent yet

Kominek closes on a note that's unusual coming from someone building an AI matching workflow.

Most professionals still don't know what's achievable with plain deterministic automation. It's reliable, it's well understood, and it handles a great deal of what people are now reaching for agents to do. Their crawl-walk-run is: learn automation properly first, because every skill required to build good automation is a skill you'll need for agentic work anyway. Then feed AI into those deterministic workflows and the results get genuinely remarkable.

Their advice to a younger version of themself is the same discipline applied to people rather than systems: slow down to go fast. When you can already see the finished thing in your head, it's easy to forget nobody else can — and taking colleagues from A to B to C is the work.

The throughline for partner teams

Strip out Zapier's specifics and Kominek's playbook makes one argument. AI doesn't remove the hard parts of a co-sell motion. It relocates them, and teams get into trouble by automating before they've noticed where they went.

Executive alignment still decides whether sellers engage. Judgment still has to happen — upstream, at validation, instead of downstream, at routing. Data quality becomes the bottleneck the moment routing stops being manual. And adoption still gates every revenue number you'd like to report.

"AI doesn't solve for those original barriers that have always sort of been there." — Dorian Kominek, Channel Leader, Zapier

That's the same shift WorkSpan is built to operationalise.

WorkSpan puts governed, partnership-specific intelligence in front of sellers inside the CRM where they already work — so the right partner reaches the right opportunity at the right moment, without depending on a partner manager holding the directory in their head, and without asking an AE to adopt one more tool. It's how partner teams get coverage at scale while keeping the specificity that makes a partner recommendation worth acting on.

If your team is working through the same validation and adoption problems Kominek describes, see how WorkSpan operationalises it. [Request a demo]

-----

Dorian Kominek is Channel Leader at Zapier, where they built the company's channel enablement and partner portal infrastructure and are now building its co-sell lead-passing motion, following a decade as an entrepreneur before moving into tech. Thanks to Dorian for sharing their playbook with WorkSpan.

‍