What Aftermarket Operators Are Getting Wrong About AI in 2026
I spent the bulk of my career in automotive aftermarket and adjacent vertical SaaS. Most recently as Vertical President, where I directed a portfolio of SaaS companies across the automotive vertical, with full P&L responsibility. Before that, in product and engineering leadership across the same space.
The aftermarket industry is now in an awkward middle phase with AI. Most distributors have heard the pitch, sat through three or four vendor demos, and signed up for a couple of AI products that looked good in a slide deck. Few have shipped anything that's moved a meaningful KPI.
The patterns I'm seeing aren't unique to aftermarket — but the way they show up in this industry is shaped by data that's idiosyncratic, operations that are blue-collar in the best sense, and a buyer profile that doesn't have an in-house tech leader to filter the pitches.
Here's what's actually going wrong, and what's worth doing instead.
ACES/PIES Is the Wrong Place to Start
Every aftermarket AI conversation I have eventually arrives at the same place: cataloging. Specifically, using LLMs to normalize, enrich, and map vendor parts data against ACES and PIES standards. Match fitment, fill gaps, deduplicate against the master catalog.
This is genuinely useful work. The data is messy in ways that take human catalogers thousands of hours per year to clean. The economics of automating it are real.
But it's the wrong place to start because it's the highest-stakes data in the business. A wrong fitment match doesn't just cost you margin — it costs you a customer relationship and an order return. The cost of a 92% accurate cataloging system at scale is higher than the cost of the manual process it's replacing, once you factor in the disputes, returns, and customer service cycles.
The right starting place is data that's high-volume, low-stakes, and currently unautomated. Sales call summarization. CRM hygiene. Vendor RFP analysis. Internal-facing reporting. These are 80%-accurate-is-fine domains where the win is real, the failure cost is low, and the team builds confidence in the tools before you point them at the catalog.
Cataloging comes after you've earned the right to point AI at your highest-stakes data. Most operators are skipping that step.
"We'll Buy a Vendor" Is Almost Always the Wrong Answer Right Now
The market is full of AI products targeted at aftermarket distributors. Some are good. Most are companies that raised in 2024, built their first integration in 2025, and don't yet have the production scars that come from running this work for real customers at scale.
The pattern I'm seeing: an aftermarket CEO buys a vendor product because the internal team can't build it, the demo looks impressive, and the contract is structured as annual upfront. Six months in, the vendor has been acquired, sunsetted, or pivoted. The integration that was "production" is now an orphaned dependency. The team still doesn't have the knowledge to fix it.
The honest read on most aftermarket-targeted AI vendors right now is that they're going to consolidate hard in 2026-2027. The ones that survive will be the ones that are either (a) backed by serious enterprise distribution, or (b) built on top of a workflow most distributors are already using.
If you're going to buy, buy the second category. If you're going to build, the cost of building has dropped enough that "we'll build it ourselves with the right partner" is now genuinely competitive with buying — and the resulting system is yours, not on someone else's runway clock.
Sales Reps Are Not the Right First AI Workflow
Every distributor I talk to wants to start with sales rep enablement. AI for prospecting, AI for follow-up, AI for CRM updates. The pitch is intuitive — sales is where revenue lives, AI saves rep time, more selling happens.
The reason this almost always fails: sales reps in aftermarket distribution are more relational than transactional. The customer is a shop owner or fleet manager who's been buying from the same rep for fifteen years. Adding an AI layer that intercepts that relationship — automated emails, AI-drafted check-ins, automated follow-up cadences — degrades the relationship long before it improves productivity.
The first AI workflow that almost always wins for an aftermarket distributor is internal-facing. Catalog enrichment for the internal cataloging team. Sales call summary feeding internal CRM hygiene without the rep having to type anything. Inventory and demand forecasting that surfaces explanations the team can act on.
Customer-facing AI comes later, and only after you've earned operational trust internally. The distributors that get this order wrong burn customer relationships that took twenty years to build.
Forecasting Without Explanation Doesn't Stick
Demand forecasting is the AI use case with the highest ROI on paper for aftermarket distributors. It's also the one that most often gets piloted and never scales.
The reason isn't model accuracy. The models are fine. The reason is that the people who would act on the forecast — buyers, planners, regional GMs — don't trust a number they can't interrogate. A forecast that says "order 240 units of part X next month" without context is a number that gets overridden by an experienced buyer who has a hunch the model can't see.
The fix is to pair the forecast with an LLM-generated explanation that sources the underlying signals: which historical pattern is driving this number, what's changed in recent months, what assumptions the forecast is making. Now the buyer has something to push back on, agree with, or override with their own intuition. The forecast becomes a starting point in a conversation, not a directive.
I've built versions of this exact pattern, and the difference between the projects that scaled and the projects that died was always whether the explanation layer was there from day one. It's the difference between AI as a tool and AI as a black box.
The Operator Who Hasn't Hired a Tech Leader Is Not Going to Hire One
The honest read on the aftermarket buyer right now: most $10-$50M operators don't have a CTO, won't hire one in 2026, and arguably shouldn't. The economics of a full-time CTO at this scale don't work — $400K loaded cost for a leader who's underutilized in a non-technical organization, with no internal infrastructure for them to lead.
What these operators need is fractional senior leadership that bridges the technology and business sides of their company. Someone who can sit in front of the board and explain the AI roadmap. Someone who can write the build-vs-buy memo on the latest vendor pitch. Someone who can own the technical evaluation work that nobody else in the building can do.
The middle market is going to consolidate around fractional CTO + AI Lead arrangements over the next eighteen months. The operators who get there earliest will lock in their advantage. The ones who keep waiting for the right full-time hire will keep falling behind, because the market for senior tech leadership willing to take a non-tech mid-market role is small and getting smaller.
What I'd Actually Tell a Distributor in 2026
If I'm sitting across from a $25M aftermarket CEO this week, the conversation I'd run is:
Don't buy any of the vendors you've been pitched until you have someone who can credibly evaluate them. Start with one internal-facing AI workflow that's high-volume and low-stakes — call summarization or catalog enrichment in the cataloging team specifically, not the customer-facing data. Build it in a way you own, with a vendor abstraction layer that lets you swap models. Write the playbook for the team that will operate it. Then expand from there.
That's not a slide deck. It's a multi-quarter engagement that builds capability without burning trust. It's also not a service most consultants are equipped to deliver, because most of them don't know what ACES or PIES is and are pitching off-the-shelf horizontal AI tools that don't fit the data shape of this industry.
If you're operating in aftermarket and want to talk through any of this against your specific situation, let's talk.