Toast's AI numbers count questions asked, and November 3 is where the dollars should appear

Toast's own AI usage report counts questions asked, not completed work, and its worker survey shows a workforce that is willing but price-sensitive. November 3 earnings are where the AI story has to show up in dollars.

Richard TangRichard Tang · 3 min read
Share
Aman Narang, Toast's co-founder and chief executive
1 / 6Slide 1 of 6
Aman Narang, Toast's co-founder and chief executive. Toast reports third-quarter results on November 3, where the company's AI pitch has to show up in dollars.

Toast−2.42% — Toast, down 2.42 percent today sells restaurant owners an AI platform, and the owners are buying: nine in ten operators use AI and 87% are comfortable with it, and Toast expanded the pitch on October 6 with Toast IQ Team for labor planning plus an expanded Toast IQ Grow for marketing 346. Before Toast reports Q3 earnings on November 3, an investor should ask one question about that pitch: how much of it is work done, and how much is people asking questions 4?

Toast counts the questions, not the work

Toast's own usage data answers cautiously. Its 2026 Restaurant AI Report counts prompts entered into Toast IQ, the AI assistant built into its platform, and the methodology states plainly that it "does not measure whether any resulting actions were completed" 3.

The composition of those prompts shows the gap. Sales and revenue questions are the largest category at 30% of threads, 53% of prompts land during peak lunch or dinner service, and 54% start on a mobile phone, which the report reads as a manager on the floor, mid-shift 3. A question typed during the rush is intent, not a completed schedule or a corrected payroll run. Prompt counts cost Toast nothing to generate, and that is why they cannot be tied to revenue without the earnings release saying so.

The workers are mostly willing, but they have their own price

Toast's blind survey of 702 restaurant employees, published October 22, sized the people who would actually use the product: about 20% are extremely open to AI at work, another 30% very open and 27% somewhat open, against 10% who reject it outright and 12% who will not accept AI training or coaching 12. A workforce that is mostly willing to let software take the busywork is exactly the adoption condition Toast needs, so the demand-side risk is smaller than the AI-skeptic framing suggests.

The pressure runs the other way too. Poor hourly pay is the top reason employees say they would leave, named by 31%, while 67% are satisfied with their pay, up 12 points from last year 12. Workers tolerating AI-assisted scheduling and upselling are being paid to accept it, which supports adoption now but leaves pay expectations to pressure operators' margins, the same margins Toast's AI is marketed to protect.

Most restaurant workers are open to AI at work, but only a fifth are eager

0102030Very open30Somewhat open27Extremely open20Not very open12Not at all open10
Data
Value
Very open30
Somewhat open27
Extremely open20
Not very open12
Not at all open10
Share of 702 U.S. restaurant employees surveyed blind by Toast in February and March 2026, margin of error about 4 percentage points. Source: Toast 2026 Voice of the Restaurant Employee Survey.1,2

November 3 is the checkpoint

Who closes the gap between asking and doing is where the revenue case lives: turning mid-service prompts into completed actions inside one platform is what Toast IQ Team claims to do, and the outcomes in its press release, like fewer overtime hours, are company-reported customer figures, not audited segment revenue 6.

So read November 3 as the test 4. Toast trades at $32.71 against a 52-week high of $39.67, and Truist+0.63% — Truist, up 0.63 percent today kept its Buy rating but cut its price target from $39 to $38 on October 19, so the AI pop has cooled 5. If the release puts dollars on the AI suite, the premium multiple has a leg to stand on. If it does not, the valuation rests on core point-of-sale and payments growth alone.

Over half of AI prompts land during peak service, when answers can still change the night

0204060Peak lunch/dinner53Morning prep19Nightly close15Mid-day lull13
Data
Value
Peak lunch/dinner53
Morning prep19
Nightly close15
Mid-day lull13
Share of Toast IQ prompts by time of day, Q2 2026 aggregated and anonymized usage data, U.S. locations. Source: Toast 2026 Restaurant AI Report via Restaurant Dive.3

Get Applications by email, free

How AI is changing the software people and companies use. In our daily email, every weekday.

Loading TOST

Reader comments

0 comments

    More in Applications

    All in this topic →

    Applications

    Podium says its AI agents hit $150 million in revenue, and its next target is ServiceTitan's contractors

    The private Utah company says its AI-agent business reached $150 million in annual recurring revenue in 2.5 years and turned profitable. Incumbent ServiceTitan, whose contractor market Podium is entering, cut operating-income guidance 33.6% on September 8 to fund its own AI platform and lost about 30% of its value.

    Applications

    Disney bids $20 million for bankrupt Character.ai's code and $180 million for its engineers

    Sworn filings show the 12-million-user companion app earning $5.3 million in August against $13.3 million of monthly cost. The model that survives is the one that meters the message.

    Applications

    ElevenLabs launches AI music tool days after a publisher priced unlicensed training at $150,000 a song

    The voice-AI company launched Instant Soundtrack on October 26, five days after a publisher priced unlicensed AI music training at $150,000 a song in federal court. Licensing is now ElevenLabs' moat and its margin bill.

    Applications

    Pinterest opens ad inventory to outside buyers via Magnite three days before Q3 earnings

    Pinterest has thrown open its display and video inventory to outside buyers with Magnite as the toll-taker, the only new revenue lever disclosed before a Q3 print where growth was guided down to 13% to 15% from 18%. The question investors must price: new incremental demand, or margin handed away at scale.

    Applications

    Perplexity lands in Hyundai cars, UK government servers and Macs while six publishers sue

    Between the close of expert discovery and a jury trial, Perplexity shipped its answer engine into a Hyundai dashboard in India, UK government and defense servers, and a free local Mac model. The reported $30 billion round funds channels with no disclosed revenue while the engine that earns the money sits in court.

    Applications

    Omnicom spent $3.2 billion on buybacks while its operations made $37.8 million

    A year after closing the IPG merger, Omnicom calls the integration substantially complete and points to 7.2% organic growth. That growth is measured against a rebuilt prior-year baseline, and the same filing shows the buyback dwarfing the cash the business produced.

    Get the newsletter

    The day's AI and market news, in your inbox every weekday.