Talkspace Inc. (Nasdaq: TALK) could compete in the wide health tech ecosystem as a provider of AI tools on top of its present focus as a digital outpatient mental health provider.

CEO Dr. Jon Cohen left all options open when it comes to how Talkspace’s soon-to-debut foundational large language models (LLMs) would fit into the company’s strategy. When asked about the prospect of licensing its tech tools during the company’s second-quarter earnings call, Cohen responded, “I would say any and all of the above, depending on what happens.”

“Once the model is built, we have certainly described internally, at least eight or 10 major possible use cases,” Cohen said. “Beyond that — whether or not someone would come to us to license it, someone would come to us to partner with it — there are some other use cases that I’m sure will evolve that we haven’t thought about yet.

“All of that is possible and on the table. The plan right now, though, is to build it and then begin to see what happens after that occurs.”

The Talkspace LLMs would be the basis for developing new clinical and administrative tools. Such models will be based on millions of documented therapeutic interventions over Talkspace’s 12-year history and refined by clinicians working on the project.

Early models are expected to debut later in the year.

Once up and running, the LLM will eventually become “an integral part of how we provide higher quality care to our Talkspace members,” Cohen said.

Specific tools may include enhanced member risk assessment and tracking, integrated mental health data in primary care clinical workflow, intake and care routing protocols, and patient engagement tools.

“Our foundational models will just power the next generation of features on our platform, but will unlock an entirely new ecosystem of applications for mental health,” Cohen said during the call.

The company has already made a handful of investments in AI tools. In March, Talkspace announced the rollout of “Talkcast,” an AI-generated podcast for individual patients that goes over key themes from their interactions with their therapist.

On the call, Cohen said that members who listen to their Talkcast go on to a second session 18% more often and a third session 23% more often than those who do not.

During the second quarter, Talkspace launched a “smart-evaluation” tool that allows therapists to forgo intake documentation in the first session and focus on trust and relationship building with patients during their first session.

Talkspace also retrained its suicidality detection model on more recent data to better understand recent evolutions in language. Cohen said the model is now 92% accurate compared to its previous accuracy rate of 83%. The company is expanding its detection tools to screen out neglect and substance use disorders.

The investments in artificial intelligence and new marketing initiatives for recently expanded Medicare Advantage and military engagements contributed to deeper losses in the quarter than in the previous year.

The company’s net loss deepened by 14% to $541,000. On an adjusted basis, earning nearly doubled to $2.3 million in the second quarter. Cash and cash equivalents were down 29% for the second quarter compared to the end of 2024.

At the top line, revenue grew about 18% to $54.3 million. Insurance-covered therapy sessions, Talkspace’s now-core and growth-driving business, increased 29% during the quarter, totaling about 385,100. Revenue from that category of care increased 35% to $40.5 million.

Revenue from direct-to-enterprise services, care delivered with contracts directly with employers or municipal entities, remained relatively flat, only slipping by 2% year over year. Ian Harris, Talkspace’s CFO, said that several deals in this category that were expected to close in the second quarter landed just after the quarter’s end.

Direct-to-consumer revenue, once the company’s bread and butter, diminished 32% in the quarter to $4.4 million.

A human in the loop

Cohen was asked for his reaction to state laws prohibiting certain uses of AI in mental health. Specifically, the analyst on the call cited Illinois banning AI in therapy. Gov. J.B. Pritzker signed the ban into law on Aug. 1.

He said Talkspace doesn’t support the use of AI as a form of therapy. But he did say that any AI-backed, mental health-related tool needed to be supervised by “licensed and real clinicians,” and be hosted and operated in an environment that comports with patient information security provisions detailed in the Health Insurance Portability and Accountability Act (HIPAA).

“Those two issues have not been addressed yet, but we do believe, as one of the leaders of the space, that those should be probable requirements as people move forward to deliver care in any manner whatever AI-mental health support is,” Cohen said.

Cohen has struck a more explicit tone in the past, ruling AI out of the primary clinical support role.

“We have no intention of using AI to provide care as an AI bot,” Cohen said during a BHB webinar. “We’re taking that off the table.”