Yelp Bet $271 Million on Hatch. The AI Battle Is Bigger Than the Phone.

The phone is one front. The customer journey is the prize.

A quarter-billion-dollar signal about where the value sits

On September 10, Hatch announced that it had integrated OpenAI’s GPT-Live-1 into its Voice AI product. OpenAI released the model in its API the same day, bringing conversations that can accommodate listening, speaking and interruptions at the same time.

The investment behind the announcement is substantial. Yelp’s first-quarter SEC filing records its February 2 acquisition of Hatch for approximately $271.2 million in cash. The filing separately identifies $30 million of employee compensation tied to future service, outside the purchase consideration.

The strategic implication is more interesting than the acquisition arithmetic: helping a contractor find a lead and helping that contractor convert the lead are different businesses. A platform that participates in both can reach farther into the customer relationship.

Hatch’s broader offering already includes multichannel agents and follow-up automation; this week’s news advances its voice layer. The important question is therefore not which vendor has discovered another communication channel. It is which approach connects those channels to the work a contractor actually wants to win.

What changes for contractors

• Evaluate the whole path from first inquiry to completed work—not just the quality of the phone demonstration.
• Treat useful replies, qualified requests and confirmed appointments as steps in one journey, with completed revenue as the destination.
• Compare the workflow and economics. TradeVulcan also uses OpenAI; the technology provider alone does not decide the product competition.

The overlooked gap is between interest and commitment

Imagine a customer reaching a replacement-service page with three questions: Do you work here? Can someone come after work? What happens during the estimate?

A phone number asks that person to switch channels. A long form asks for a commitment before the questions are answered. A helpful conversation can resolve the uncertainty first and make the next step easier.

Now move the same customer forward a week. The estimate is in their inbox. The next conversion may not require another sales presentation. It may require explaining an option, confirming a scheduling detail or getting the right person to answer a concern.

These are different points in the sale, but they share the same operating problem: customer intent exists, and something stands between that intent and action.

That is the opportunity for agentic AI—software that can interpret the request, choose a relevant next step and carry out an authorized action. The objective is not a longer conversation. It is a more useful one.

OpenAI is the engine. The application still has to drive the job.

TradeVulcan also uses OpenAI in its AI workflows, including Site Spark’s conversational assistance and intent handling. Access to capable models is part of the foundation; the contractor-specific application is where the business problem gets solved.

OpenAI’s GPT-Live-1 architecture makes that distinction visible. Developers choose the backend models and tools behind the conversation. Natural speech is one layer. Looking up information and taking action are additional layers.

For a home-service business, those actions need operating context. The system needs to distinguish a repair request from an estimate, understand the service area, recognize urgency and move the customer toward an appropriate booking or human handoff.

The competitive advantage has to be built into those decisions and connections. Two products can use the same AI provider and still offer very different customer experiences, implementation paths and economics.

The useful question is not simply “Whose AI is this?” It is “What can this system get done for my customer and my team?”

Microconversions are the small yeses that make the sale possible

A microconversion is an intermediate commitment that moves the customer closer to service. Providing a usable contact number, confirming a service need, replying to a text and accepting an appointment can all advance the opportunity.

AI can help lead those steps by asking the relevant question, removing uncertainty and offering a next action that fits the customer’s readiness. Someone who wants to keep texting should not have to start over on the phone. Someone ready to book should not be left with a generic promise of a callback.

The distinction is between activity and progress. An opened chat is activity. An eligible service request with enough information to act is progress. A delivered follow-up is activity. A customer reply that gets an estimate moving again is progress.

Thinking this way changes what a contractor asks software to do. Instead of buying another channel and hoping the office connects the pieces, the business can design the transitions that keep the sale moving.

Where the next conversion happens

Small improvements can compound without buying another lead

Consider a hypothetical monthly funnel with 1,000 eligible inquiries. If 60 percent become useful two-way conversations, 60 percent of those conversations become bookings and 85 percent of bookings become completed jobs, the result is 306 completed jobs.

Raise the first rate to 70 percent and the booking rate to 65 percent, while holding the number of inquiries and completion rate constant, and the result becomes approximately 387 jobs. That is about 81 additional jobs, or a 26.4 percent increase, from the same starting demand.

This is illustrative funnel math, not a vendor performance result. Its value is showing why more than one conversion point deserves attention. Better engagement and better booking can reinforce each other; neither requires pretending every conversation is a sale.

Illustrative model: the same 1,000 inquiries, a different result

TradeVulcan’s bet is on the connected journey

A product for the point where the opportunity stalls

The handoff is part of the product

A contractor should not have to choose between automation and an accountable office. A good handoff makes both more effective: the customer keeps moving, and a person receives the service need, conversation history and next action instead of another unexplained notification.

TradeVulcan’s booking and CRM actions depend on the connected system, permissions and selected workflow. Site Spark can preserve a lead when a booking cannot finish; Close Forge provides stop conditions and human ownership around follow-up. Those mechanisms support the same commercial goal: move the opportunity forward without creating cleanup for the team.

That is also a useful test for the entire category. The glamorous part of a demo is the conversation. The consequential part is what appears in the office afterward.

Does the right person know what to do? Is the appointment real? Does the follow-up stop when the customer responds? Does the customer have to repeat the story? These questions reveal more about a product’s operating value than its ability to deliver a smooth greeting.

The economics have to work for the contractor, not just the software company

The price of an acquisition says something about the buyer’s strategy. It says nothing about what an individual contractor should pay to improve conversion.

TradeVulcan’s published monthly plans start at $299 for Starter, $699 for Growth and $1,299 for Pro, with two, three and five included marketplace apps respectively. Additional apps are available as add-ons. The plans advertise a 14-day trial and monthly billing without a long-term commitment.

That creates a practical alternative for cost-conscious operators: choose the part of the journey that needs attention, keep a supported existing CRM and add workflows as the business case develops. A smaller contractor does not need to approach every conversion problem as a wholesale technology replacement.

Hatch publishes an annual, per-location agreement structure with a platform fee plus usage and provides dollar pricing by quote. Compare the same channels, locations, usage and setup scope before calculating savings.

The stronger value question is how much incremental contribution remains after the complete software bill. That includes the selected apps, usage, setup and staff time needed to operate the system—not just the entry subscription.

The scoreboard should end with completed work

Channel reports can make the same opportunity look like several wins. A web chat captures the lead, a text helps book it and a call confirms the visit. The company still has one customer and one job.

Measure those interactions as contributions to a single journey. Preserve the source, distinguish a requested time from a confirmed appointment, and connect the opportunity to completed work and collected revenue. Use microconversion rates to find the weak step rather than adding them together as if each were a separate sale.

Then ask what improved because of the system. Compare similar lead cohorts or use a controlled rollout where practical. More bookings are helpful only when they produce the right work at an acceptable cost; additional cancellations, poor-fit dispatches or staff correction time belong in the calculation.

For operators, this is the shift from buying communication capacity to buying accountable execution. For software companies, it is a more demanding contest than producing the best-sounding demo.

Make the next AI demo follow a real customer journey

• Begin with a website question, then move to text. Check whether the service context follows the customer.
• Ask for an appointment outside the service area or available schedule. Look for a useful alternative, not an invented promise.
• Reply to estimate follow-up and request a person. Check that the automation stops appropriately and the team receives the history.
• Trace the customer into the CRM and the eventual job outcome. Price the complete workflow against the additional completed work it would need to earn.

The next customer may never call

Yelp’s investment in Hatch is a major bet on what happens after demand is created. The new voice technology is an important part of that story. It is not the boundary of it.

A homeowner can become a customer through a web conversation, a text exchange, a guided booking or the answer that finally gets an estimate approved. The larger agentic-AI opportunity is to connect those moments, using context to choose the next useful action.

That is the market TradeVulcan is building for: voice, SMS, website conversion and follow-up, supported by OpenAI and a modular commercial model. Its case to contractors rests on the journey it can support and the economics of putting it to work.

Answering the phone is valuable. Helping the homeowner take the next step—wherever the conversation starts—is the bigger prize.

The questions behind the next AI purchase

Updated September 12, 2026

Expanded from the original voice-AI report to examine cross-channel lead conversion, TradeVulcan’s approach and contractor economics. The original publication date and URL are preserved.

Methodology

Acquisition terms are sourced to Yelp’s SEC filing. Product and pricing information was reviewed against Hatch, OpenAI and TradeVulcan materials on September 12, 2026; TradeVulcan’s OpenAI use was also checked against its Site Spark implementation. Product scope and direct CRM actions vary by plan, connector and configuration. The funnel table is an illustrative calculation, not observed customer performance or a head-to-head benchmark: 1,000 × 60% × 60% × 85% = 306 expected jobs; 1,000 × 70% × 65% × 85% = 386.75, a 26.4% increase. Customer scenarios and buying criteria are Dispatch analysis. The hero is an owned sample product screenshot.