AI Belongs Where the Work Happens
Salesforce is spending billions to put AI closer to customer data and everyday workflows. Contractors should pay attention.
Salesforce is reportedly considering spending $2 billion on a three-year-old company that uses AI to help businesses understand what their customers are thinking.
On its own, that's another wild AI valuation story. Listen Labs runs AI-moderated interviews with customers at a scale no human research team could match, then turns the raw conversations into a report you'd normally wait weeks for, in a matter of hours. Two years ago the company didn't exist. In January, it was valued at $500 million. More recently, it reportedly signed a term sheet for a $125 million round at a $1.5 billion valuation, then walked away from it while acquisition talks with Salesforce were underway. Salesforce is reportedly discussing something closer to $2 billion. Nobody has confirmed a deal. It may not happen at all.
But look at what Salesforce has been buying for the past two years, not just this one deal, and something more interesting starts to show up.
What Salesforce is actually buying
Salesforce has picked up a small pile of AI companies recently, and the pattern isn't really about what any single one of them does. It's about where Salesforce keeps saying it's putting them.
Convergence.ai builds agents that can navigate a web interface and carry out a multi-step task on their own. Salesforce folded it directly into Agentforce. Bluebirds does AI-driven sales prospecting. Salesforce built it into Sales Cloud. Regrello turns messy business documents into automated workflows. Salesforce wired it into Agentforce and Slack. Momentum takes the unstructured mess of a sales call and turns it into something usable, and it feeds straight into the same two places. Informatica, the largest purchase of the group at a reported $8 billion, isn't a feature at all. It's meant to become the data plumbing underneath everything else Salesforce is building.
Read enough of these announcements and one word keeps showing up: integrate. Not launch. Not introduce. Integrate. Salesforce isn't building a new place for people to go ask AI questions. It's building AI into the places people are already working.
And that's really the whole idea, stated as plainly as I can put it.
The future probably isn't your team "using AI." It's AI becoming part of the work they're already doing.
That's an important distinction. Most businesses, yours included, don't need another place for someone to go ask an AI to do something for them. They need the phone call summarized without anyone typing notes. The customer record updated without someone copying information from one system into another. The estimate that gets followed up on without a CSR having to remember it's been sitting there for two weeks. The customer from three years ago who resurfaces because there's an actual reason to call, not because someone finally got around to a list.
AI belongs where the work happens.
What this actually looks like
None of this requires a business to have a plan for AI. It requires the software already running the business to get better at the parts nobody enjoys doing anyway.
A call comes in, and the system already knows who's calling and what they had done last time, before anyone picks up.
The conversation gets summarized without a technician typing notes on the way to the next job.
An estimate goes out, and it follows up on itself instead of waiting for someone to notice it's gone quiet.
A customer from three years ago resurfaces because there's an actual reason to reach out, not because someone finally got to that list.
A review request goes out the day the job closes, without anyone remembering to send it.
Nobody has to retype anything from one system into another.
None of this is dramatic. That's the point.
Home service software is already fighting over this
This isn't hypothetical for your industry. It already happened.
In August 2026, ServiceTitan cut off Podium's integration into its platform, ending a partnership the two companies had run for nine years. About a thousand shared contractor customers got thirty days' notice.
The apparent reason: Podium's AI agent had gotten capable enough to answer calls, update customer records, and book jobs directly, moving Podium from a communications layer sitting on top of ServiceTitan toward something that looked a lot more like a competitor. The hard evidence is in ServiceTitan's own developer terms, which were updated to specifically bar a partner's AI from independently deciding what data to touch or what actions to take on a customer's behalf.
Contractors caught in the middle reportedly lost synced customer records, automated texts, and review requests, in the middle of running their businesses, because two software companies were fighting over who owns the workflow.
That's the clearest evidence for everything this article has been arguing. Not a feature announcement. A breakup. Once AI gets capable enough to actually do the work instead of just assist with it, the question stops being who has the coolest AI and starts being who controls the place where the work happens.
Built in is different from bolted on
Here's where it gets useful for you specifically.
ServiceTitan has real, sophisticated AI, but its most capable features live inside a separate "Pro" product, layered on top of the platform contractors already pay for. One ServiceTitan customer put it plainly in a public review: they wanted the AI "fully functional as a core component, not as a Pro product." Housecall Pro's calling AI is its own separate subscription too, though to the company's credit, its documentation says plainly when a feature isn't live yet instead of pretending it is.
Jobber is doing something different. Its AI is available on every plan and reads a customer's actual job history without a separate login. Still in beta, it also lets a technician create a job, write up a quote, and log field notes by voice, from the truck.
That gives you a real test, and it's a simple one: if getting the AI requires another login, another dashboard, another integration, and another monthly bill, the problem probably isn't solved yet. It's just been renamed. The interesting version of this doesn't ask anyone to remember to use it. It's just part of how the job gets done.
Where this can go wrong
None of this means embedded AI is a sure thing.
Salesforce reports Agentforce annual recurring revenue above $1.5 billion, growing more than 240 percent year over year. That's significant. But it isn't the same thing as proving customers are getting $1.5 billion worth of value from it. McKinsey found that only about a fifth of organizations using generative AI had fundamentally redesigned at least some workflows around it, while more than 80 percent reported no tangible impact on enterprise-level EBIT.
And putting AI closer to the work creates its own problems. AI acting on bad customer data doesn't produce a good outcome faster. It produces a bad outcome faster. Giving one software company more control over your data, communication, and workflows can also make switching later considerably harder.
So the honest version isn't "embedded AI is here, get on board." It's that this is clearly where the money and product development are headed. Whether it actually makes the contractor's business better is still the part that matters.
What you should actually do about any of this
Nothing dramatic. That's kind of the whole argument.
Don't go buy an AI tool because you read this.
Look at the software you're already running your business on. What's becoming native? What's another subscription wearing the word "AI"? Does it use your actual customer and job data? Does it take action, or does it just generate text somebody still has to act on?
And before any of that: what's the actual business problem you're trying to fix?
That's the question that matters.
If you want the fuller picture of where AI actually helps a contracting business, we've laid it out separately.
AI isn't the strategy. It's one way we fix the leak.