Over the past year, AI phone calling has leapt from Hayes Valley hackathon project to remarkable market traction. Today, most of this revenue is concentrated in upmarket customer service organizations. This concentration is reflective of the historic implementation demands of AI calling and also of the product approaches that venture dollars have flowed towards to date. It is not, however, reflective of a demand that only exists in the enterprise. AI calling feels like magic, and customers across verticals and segments are grasping for products purpose-built to their use cases.
Given this underserved downmarket demand, AI calling is emerging as a powerful entry point into myriad verticals historically resistant to software. What’s more, this lightning-in-a-bottle capability is broadly available to entrepreneurs; while our portfolio company Bland does focus on selling to enterprise customers directly, the company also empowers developers to build specialized, vertical-specific solutions atop the company’s Pathways platform.
That AI calling is both desperately desired by customers and available as an off-the-shelf capability presents an interesting set of market selection constraints to entrepreneurs. In markets where this is the right wedge, the entrepreneur must also consider adjacent soon-to-be competitors and, in many cases, plan to be ‘more than’ AI calling before long.
Below are some of the key characteristics for new vertically-focused AI calling companies to thrive. We are keen to invest in categories where all four conditions are met.
1. There is demand driven by call center hell, or a phone that rings with the sound of money.
Calling operations should be both critical and painful for the company’s prospects.
It is customers who already manage a call center, or who scramble to answer the phone when it rings, who grasp for the AI-calling-painkiller when it is offered. (A good example is home services, where Avoca and Sameday focus, an industry in which 80% of revenue is booked over the phone. Another is loan servicing, where Salient and Kastle play.)
On the other hand, buyers with hypothetical use cases for AI calling, where no human to be replaced exists, tend to show interest but not urgent demand. It is harder to run fast with such prospects.
2. Competition with vertical and function-specific incumbents must be winnable.
It’s much harder to succeed when there is an agile and established incumbent for whom AI calling is an obvious upsell product.
Demand for AI calling does present the perfect wedge with which to challenge many a flat-flooted incumbent or stalled growth stage company. (Toma is taking advantage of such an opportunity in auto shops, where no cloud native shop management system has yet gained critical mass. There is a similar opportunity in software for local government, where Polimorphic is focused.) On the other hand, because AI calling is now an off-the-shelf capability, we expect that capable incumbents, like ServiceTitan, will add the feature where it’s relevant.
Regardless of the incumbent’s agility, the mitigant to this kind of competition is the same: run fast and establish a big footprint before the incumbent catches up on product. The investment judgement to be made is one of, “how far can the startup get before that happens?”
3. Competition or frenemyship with horizontal AI calling platforms must be sustainable.
Founders should have a clear answer as to why larger prospects will pick this solution over buying from the horizontal AI calling platforms (which their vertical solution may be built atop!).
Some examples of paths to sufficient differentiation:
- Lots of market size downmarket, as small ACVs do not merit implementation engineers, and self-serve construction of calling pathways can be challenging.
- Vertical-specific demand for specialized features, as these will likely fall off the roadmap of horizontal providers, at least for time to come.
- Bundling with a suite of vertical specific products, so long as there are enough jobs to be done for the customer.
An example: HelloPatient’s singular focus on healthcare means they differentiate through productizing deep integrations with painful industry-specific systems and processes.
4. There is lots of available roadmap through which to build a moat.
When leading with a wedge that offers no defensibility, the key to success is running fast on roadmap, and leveraging lightning-in-a-bottle initial demand to fuel rapid expansion into more areas of the customer’s business.
It is through such product expansion that a commodity wedge in “AI calling for X” can become a defensible business of scale. Such product expansion can also enable a previously small market to become a significant one through much larger ACVs.
An example: Zingage, which sells to home nursing agencies, leads with AI calling for scheduling, but is rapidly advancing to be the “AI secretary for home based care.”
Where there is room beyond this framework
Looking for one-to-one replacements of tasks as done by humans can be a useful heuristic for demand, but it is myopic of the latent demand for new ways of doing things that AI can address. There are several compelling opportunities for Voice AI where the above framework does not apply. These use cases are largely hypothetical ones (in that dedicated call centers do not serve the demand today) in which there is massive latent demand nonetheless.
Two notable examples:
- Interview-driven data collection at unprecedented scale: There are many kinds of data that are best collected through voice conversations. Some are concentrated in labor markets, like candidate screening. Others include market research and polling.
- Voice AI for companionship and care: Voice AI is uniquely positioned to address growing needs in caregiving (mental health, or memory care), tutoring (making 1:1 teaching more accessible), and companionship (addressing loneliness).
