AI in the sale. What owner-led companies are doing with it, and what it's worth.September 29, 2026 · 8 stories
This week
AI multiplies whatever's already there
A property manager caught its AI agent sharing wrong information. Plus a 90-day revenue stall, speed to lead and what bad CRM data does to AI.
Editorial illustration / AI-generated
Hey,
Issue three is a little different. Instead of wins, I went looking for businesses where AI didn't pay off, and for what was sitting underneath when it didn't.
Named failure stories are hard to find, because most companies don't publish their misfires, so this week leans more on research and first-hand accounts than usual, and nothing carries the measured label. A few items are about the foundation on its own, with no AI rollout attached. In the two real stories, the AI mostly did what it was set up to do, and the trouble was in the setup around it: the information, the data and the handoffs. The research points the same way.
01
Property management
A property manager found its AI leasing agent was giving prospects wrong information. The gaps were in its own setup.
Editorial illustration / AI-generated
Northpoint Asset Management, a Salt Lake City property manager with more than 8,000 units, runs AppFolio's AI agents for leasing and maintenance. Chief Innovation Officer Steven Rea says the team found gaps in the property and policy information it had given the leasing agent by auditing the agent's conversations with prospects: “If we hadn't been actively monitoring that, we would have been burning leads that we didn't necessarily need to burn because we were sharing inaccurate information.” The second problem was that leasing staff kept stepping into the agent's conversations. In AppFolio, once a person steps in, the agent can't pick that conversation back up, at least not today. Northpoint's fix was a written rule: the agent owns first contact through tour scheduling, unless it can't answer a question or the prospect asks for a person.
His advice for anyone about to switch an agent on: describe the workflow first, including who owns the result and when the agent hands off to a person, and start with one workflow.
Anecdotal One operator's account on The Residential Operators podcast, quoted from the episode's captions. No numbers given. Northpoint is bigger than most readers' companies (8,000+ units as of 2025), but the leasing setup maps onto any inbound sale.
The part I'd copy is the audit. They found the problem by reading what the agent was actually telling prospects. If you put AI on your inbound, have someone read a sample of its conversations every week for the first couple of months, and write down who owns a lead at each step before it goes live.
02
Speed to lead
Three in four service businesses don't answer a new lead within five minutes
Editorial illustration / AI-generated
Blazeo surveyed 573 service businesses across financial services, real estate, home services, professional services, legal and healthcare. 74% said they don't respond to new leads within five minutes. Only 35.4% said a five-minute response is essential, and 62.1% of that group said their teams actually meet it, so nearly 38% of the businesses that call it critical miss their own standard. Blazeo's CEO, Ashhad Syed, puts it down to coverage and systems: “Businesses do not have a motivation problem. They have a coverage, handoff and systems problem.”
74%
Don't respond to new leads within five minutes
Vendor-reported survey, 573 service businesses
Vendor-reported A vendor survey (Blazeo sells lead-response services). Answers are self-reported, and no survey dates or company sizes are given. The reasons are Blazeo's reading of the results, and the report says it shows associations, not causes.
An AI assistant can answer a lead in seconds at 10pm, which is why it gets bought for this. It only helps if the lead then lands somewhere a person will see it, with an owner and a next step. I'd check that first: pull your last 30 inbound leads and write down when each one came in and when a person first actually spoke to them.
03
Outbound
An AI workflow replaced three SDRs and cut costs 87%. Revenue stayed flat for 90 days.
Editorial illustration / AI-generated
87%
lower cost: $18K a month in SDR pay down to $2,400 in tools
+35%
pipeline by day 90, with revenue flat
Half
the meeting-to-opportunity conversion it had before
3x
revenue per meeting after the rebuild
Asfand Rafiq on LinkedIn, one unnamed client, September 2026
Asfand Rafiq, a revenue consultant, wrote on LinkedIn about a client where he replaced three SDR seats with an automated Clay-to-Instantly workflow in the fourth quarter of last year. Cost went from $18K a month in loaded SDR pay to $2,400 a month in tools. For the first 60 days outbound volume was four times higher and reply rates held. By day 90 pipeline was up 35%, revenue was flat, and conversion from meeting to opportunity had dropped by half. His explanation: the workflow was booking calls with people who fit the profile on paper but had no active reason to buy. The SDRs had been judging timing from things like job postings and company announcements, nobody had written that down, and so nobody built it into the workflow.
Rebuilding the workflow with buying signals as the first filter took three weeks. Pipeline volume fell 40%, and he says revenue per meeting tripled.
Anecdotal One consultant's post about an unnamed client. We could only read the post itself and couldn't check it further.
I think the missing piece was simple: nobody had written down how the SDRs decided who was worth a call. Most of you don't have SDRs, but if you're the one deciding which leads are worth your time, that judgment lives in your head too. It's the first thing I'd write down before automating any part of your sale.
04
Research
Only 21% of marketers say their CRM data is very well prepared for AI. 62% say bad data already costs them revenue.
Validity surveyed 500 B2B and B2C marketers in the U.S., U.K., Brazil, Australia and New Zealand for its 2026 CRM data report. Just 21% said their CRM data is “very well prepared” to support AI, and 62% said their organization loses revenue directly because of poor CRM data quality. Nearly a third of teams spend six or more hours a week fixing and reconciling data. And nearly 78% of C-suite respondents said they had acted on an AI recommendation they later suspected was wrong because of bad underlying data, against 41% of individual contributors.
CRM data and AI
CRM data and AI
Say their CRM data is “very well prepared” for AI
21%
Say they lose revenue directly to poor CRM data
62%
C-suite: acted on AI advice they later suspected was wrong because of bad data
78%
Validity, State of CRM Data Management in 2026, 500 marketers. The last figure is “nearly 78%”, C-suite respondents only.
Vendor-reported A vendor survey (Validity sells CRM data-quality tools). It asked marketers, not salespeople, and doesn't give company sizes.
I think an AI tool reads whatever is in your CRM and acts on it without the second-guessing a person would do. So if a lot of your contacts are out of date, its answers will be confidently out of date too. I'd pick one small habit before any AI project, like every open deal having a next step and a date, and hold the team to it for a month.
05
Software
A sales leader started rolling out AI tools and found her CRM was a mess. Her best rep's deals lived in text messages.
Editorial illustration / AI-generated
Marchelle Mooney, VP of Sales at Mangomint, which sells software to salons and spas, described this on stage at SaaStr AI London, according to SaaStr's write-up. When she started rolling out AI tools, including Momentum, she found the company's Salesforce was a mess: deals closed with no notes, no call logs and no contact records. The marketing automation platform held 10,000 more data points than Salesforce, even though Salesforce should have held more. Her top rep was closing 35 logos a month, and everything lived in text messages.
10,000
More data points in the marketing platform than in Salesforce
Per SaaStr's write-up
Her advice, as SaaStr reports it: pick AI tools based on what makes sure everything flows back to one source of truth, then build agents on top of clean data.
Anecdotal One sales leader's account, paraphrased by SaaStr rather than quoted directly. From April, so older than the rest of this issue. Company size isn't given.
This is common. The best rep is often the one the CRM knows least about, because they're too busy selling to type. That's fine until you want AI to learn from how they sell, or until they leave. I'd make logging a call take one tap from their phone, so the CRM becomes the easy path for them.
06
Insurance
79% of insurance agencies use AI. 60% say their biggest problem is systems that don't connect.
Editorial illustration / AI-generated
Augie, a volunteer-led insurance technology exchange, surveyed more than 600 insurance professionals at agencies and carriers. 79% of agencies said they use AI in their workflows, and 60% named limited system integration as their most significant operational challenge. 60% of agencies said they re-enter information into several systems across carriers, and half of all respondents said quoting is still the most manual process, ahead of submissions at 39% and renewals at 36%. Augie's executive director, Cal Durland: “The industry doesn't have a shortage of innovation. It has a shortage of integration.”
50%
Say quoting is still the most manual process
Augie Connectivity Survey, 600+ respondents
Directional An industry survey of agency and carrier staff. The split between agencies and carriers isn't given.
I'd guess most owner-led companies have a smaller version of this. The quote lives in one tool, the customer in another and the follow-up in someone's inbox, and an AI tool added to the mix can easily become one more place to re-type things. I'd map where a quote actually travels, from request to signature, and count how many times someone copies the same information by hand.
07
Research
Gartner predicts sales leaders who overhaul their data and systems will be five times as likely to see a return on AI
Editorial illustration / AI-generated
In a July prediction, Gartner said that by 2028 AI agents will outnumber sellers ten to one, yet fewer than 40% of sellers will say the agents have improved their productivity. It also said chief sales officers “who overhaul data, automation and user experience will be five times more likely to gain ROI from AI than those choosing quick fixes.” Dan Gottlieb, VP Analyst in Gartner's sales practice: “If those systems are fragmented, the agents will scale the fragmentation.”
5x
More likely to gain ROI from AI after overhauling data, automation and user experience
Gartner prediction, July 2026
Directional Analyst predictions for 2028, not survey results, and aimed at large sales organizations. The related Gartner survey was of 210 chief sales officers and senior sales executives in January and February.
Five times is a forecast, so I'd hold it loosely. The quote is the part worth keeping. An agent does the same thing over and over, so if your handoffs and data are patchy, you'll get patchy results at a much higher volume.
08
Idea
Two people deep in AI sales tools say the same thing: do the groundwork first
Editorial illustration / AI-generated
Jason Lemkin, who runs SaaStr and its 20-plus AI agents, wrote in August that “AI GTM agents don't figure out your go-to-market. They multiply what's already there,” and that “10x times zero is still zero.” He says about 80% of the conversations he and SaaStr's Chief AI Officer have with companies are some version of “I bought a tool and I have no leads.” His list of what to have first starts with at least one human rep closing deals with a defined process, written down specifically enough that a new hire could follow it. St John Dalgleish, CEO of the AI outbound company Perlon AI, said on The AI Revolution Show that for most companies the answer to whether AI SDRs work “so far has been no,” and that “you've got to do some fundamental work on the business first.”
Dalgleish's other point was ownership: the companies he's seen succeed hand responsibility for it to one person, and the more senior that person is, the better.
Directional Two expert views, both from people with a stake in AI sales tools. Lemkin's 80% is his own estimate, not a measured figure.
I'd take their checklist almost word for word. If you can't write down how your best deals get won, clearly enough that a new hire could follow it, an AI tool has nothing to copy. That write-up is also the first thing I'd hand a new salesperson, so it pays off even if you never buy the tool.
The takeaway
What this means if you're still the one selling
Look at what sat underneath each of these. The leasing agent needed accurate information and room to do its job. The outbound workflow needed the timing judgment the SDRs had never written down. The sales leader's AI rollout turned up a CRM with none of her best rep's notes in it. The research says the same thing more broadly: CRM records nobody kept up, gaps and handoffs in how fast leads get answered, and systems that don't connect.
The order I'd go in: write down how your sale works today, including the judgment calls. Give every lead an owner and a next step in one place. Measure how fast a person actually responds. Then put AI on the part that's already working, and read what it does for the first few weeks.
Cheers, TeeJay Johnson Owner, Foundry Revenue Partners
Editorial illustration / AI-generated
Free pre-assessment
Is your sales process ready for AI?
The free pre-assessment looks at how your sales engine is running today against your goals, before anything gets recommended.