The 2026
Annual Letter.

To the operators we work with, the operators we hope to work with, and anyone trying to make sense of where artificial intelligence belongs in their business — this letter is for you.

I started DRS Automation Group in November 2025 because the way most businesses were being sold AI didn't match what I was seeing in the field. The pitch decks were glossy. The demos were polished. And the operators I talked to — pest control owners, dealership managers, service-company founders — were still drowning in the same problems they'd had a year before. Missed phone calls. Cold leads. Quotes that took three days to follow up on. The technology had clearly arrived. The deployment had not.

So DRS exists for one reason: to close that gap. Not by selling another chatbot. Not by writing strategy decks that gather dust on someone's desktop. By engineering the specific system that removes the specific bottleneck — and then running it.

01What we did this period.

The first builds were custom. One operator at a time. Voice agents that answer every inbound call. Lead qualifiers that close conversations in SMS. Automated dispatch that routes the right technician to the right address on the right day. We engineered them, deployed them into production, and operated them.

By the third month, a pattern emerged. The architecture that worked for one pest control operator worked for the next. The same was true for independent auto dealers. The custom builds were starting to look like a category.

So we productized them. PestPilot became our first vertical operating system — answering calls, qualifying leads, dispatching the optimal technician, booking appointments. Live in production for pest control operators. LotPilot followed for independent auto dealers — working inbound leads in under thirty seconds, qualifying buyers over multi-day SMS cadences, booking test drives directly into dealership calendars.

The custom work didn't stop. It accelerated. Most of what DRS engages on is still bespoke — every operator is different, and the productized side exists to prove the custom side compounds, not to replace it.

"Most agencies sell automations. We sell expertise. The work is the proof — and our products are the proof that the expertise holds in production."

02What we learned.

Three things have become unmistakable over the last six months.

First — vertical AI beats horizontal AI, every time. A system engineered for pest control will out-perform a generic copilot at pest control. Not because the generic copilot is bad — because the narrower the scope, the higher the precision. Pest operators care about WDIR letters, day-off scheduling, geographic routing, and termite handoff rules. None of that exists in a generic AI receptionist. All of it exists in PestPilot because we built it that way on purpose.

Second — operators are the most underserved audience in the AI conversation. Enterprise AI has consulting practices throwing thousands of slides at problems. Consumer AI has the entire venture industry behind it. The owner running a $5M pest control operation has almost no one talking to them. That's the wedge. That's where DRS belongs.

Third — the only thing that matters is whether the system runs in production. A scoped proposal is worth less than a deployed agent. A clever architecture is worth less than a real call answered at 6:47 PM on a Tuesday. We rewrote our engagement model around this. The systems we build go live before we send a final invoice. The proof is the work.

03What we got wrong.

Plenty — and I will tell you the specific ones, because in this work the only honest annual letter is the one that names its mistakes.

We shipped a routine configuration change to a voice agent's tooling on a Sunday night. The API we were sending a PATCH to interpreted partial updates as full replacements — every field we didn't include got silently wiped. Monday morning, nineteen customer calls hit a tool pointing at nothing. Zero bookings. A clean five-figure revenue miss on a single line of careless code. The bug repeated twice more in subsequent weeks before we wrapped every destructive API call in a helper that fetches the full record first. We wrote this up in detail in Note 01.

We burned our cold-email domain reputation pushing too hard, too fast, with too generic a hook. Open rates went from healthy to floor in five days. Recovery required a five-day domain warmup, a verified list, and a rewritten opener grounded in real client proof. The lesson: cold outreach is a precision instrument, not a volume instrument.

We mis-architected our first deployment with a phone-system partner — the agent went live on Monday morning and got zero calls because somewhere upstream the destination ID had been quietly reshuffled. The agent was healthy. The plumbing was unplugged. We had no monitoring for volume regressions; we built that next.

Every one of these mistakes is now operationally embedded. We don't ship destructive API calls without a safe-patch wrapper. We don't run cold outreach without a warmup discipline. We don't trust integration heartbeats without baseline-aware volume alerts. The mistakes are the curriculum.

04What we believe.

Six months in, three convictions have hardened into operating principles. Each one is the subject of an essay in the firm's research, and each one shapes how we engage every new operator who calls.

Vertical AI compounds. Horizontal AI plateaus. A general-purpose agent reaches eighty percent of any vertical's conversations in a week — then stops. A vertical agent built for one industry keeps improving forever, because every edge case the firm encounters becomes part of the system. This is the entire logic of the holding-company architecture. We build for one industry at a time, deeply. We resist the urge to be everything to everyone. (Note 02 — the thesis.)

Most "AI failures" are integration failures. The model is the easy part. The plumbing — the connection to the CRM, the calendar, the dispatch system, the phone tree — is where production systems live or die. Treat integrations as the product. Wrap every destructive call. Verify every write. Build baseline-aware monitoring, not just heartbeats. (Note 01 — the field.)

Speed of response is the most under-priced variable in operator-led sales. The data on this is two decades old and barely talked about: a five-minute response converts at roughly twenty-one times the rate of a thirty-minute response. Artificial intelligence is the first technology capable of closing that gap at scale. The operators who install it now will have built a structural advantage by the time the rest of their industry notices the asymmetry exists. (Note 03 — the economics.)

If you want a longer read on any of these, the firm publishes its thinking at drsautomationgroupllc.com/research. We will keep adding to it. The work is the proof. The thinking is how we earn the right to do the work.

05Where we're going.

The next twelve months are about three things.

More verticals. PestPilot and LotPilot are two of an eventual portfolio of vertical operating systems for operator-led industries. Several more are incubating right now. We won't ship them publicly until they're in production with at least one real operator. We don't pre-announce.

More custom depth. The bespoke side is the work and we're investing more, not less. Better diagnostics. Faster scoping. A standing engagement model that lets us deploy a custom AI system in 4-12 weeks with the same precision a productized vertical gets.

Earning the right to be the firm operators call first. Five years from now, the question "who do we trust to figure out AI for our business?" should have one obvious answer for the operators we serve. That's the brand we're building. That's what every engagement, every system, every letter like this one is in service of.

"We're not here to sell automations. We're here to be the firm that operator-led businesses trust to engineer their AI — the same way they'd trust an architect to design a building or a surgeon to perform an operation."

06A note on the long arc.

DRS is six months old. The opportunity in front of us — vertical AI for the businesses actually running the economy — is decades long. The next decade of small-and-mid-market business will be reshaped by artificial intelligence in ways most owners are not ready for. The firms that show up with engineering discipline, vertical depth, and real systems running in production will compound. The ones that show up with decks and demos will not.

We intend to compound.