Specialized AI built around proprietary data and industry workflows is becoming the main engine of growth, according to discussions at the Bank of America Private Tech Trailblazers Conference 2026 in Palo Alto. Nine featured companies, from restaurants and construction to hospitals, payments and defense, target defined problems where AI delivers measurable results. Executive analyst John Furrier said specialized intelligence was a sleeper story a year ago and is now central because systems derive intelligence from data. The shift matters for business as foundation models become interchangeable.
How nine private companies apply vertical AI
Bear Robotics has about 16,000 autonomous mobile robots in the field, a backlog of about 4,000 units and revenue that doubles every year, co-founder Bren Pierce said. Its core market is restaurants in Japan and Korea, plus care homes and casinos, while a partnership with LG extends it into warehouses and factories. Humanoids from its reacquired Kinisi startup run on the same software base and cloud infrastructure as mobile robots. Foundation models learning from a few hundred examples, onboard compute such as Nvidia Jetson Thor and coding with large language models cut work that took six months to days, though tactile hands still cost about $30,000 per hand.
Address Robotics, operating as All3, addresses construction where labor makes up 55% to 60% of costs, CEO Rodion Shishkov said. Its software designs a building from plot to permit-ready documents, a production layer directs industrial robots to make one-of-a-kind elements at mass-production cost, and its Mantis mobile robot assembles them on site. Because the company designs process and robot together, it removes most corner cases instead of training robots for each one. After a seed round of about $25 million to $30 million, it is preparing its first site for a six-story co-living building on an 11-sided plot.
Bloomreach, Harbinger, Unconventional AI and Airwallex show how vertical models, hardware and financial infrastructure scale. Bloomreach uses about 100 models, and its Loomi AI engine trained on 7 billion consumer profiles performs five to 10 times better than off-the-shelf LLMs, CEO Raj De Datta said. Harbinger sells electric and hybrid medium-duty platforms at diesel prices, saving a typical California parcel truck about $30,000 yearly on fuel after charging, CEO John Harris said. Unconventional AI raised about $540 million, grew from 14 to about 60 staff and taped out a chip at Taiwan Semiconductor Manufacturing Co. on June 1. Airwallex, founded in Melbourne in 2015, holds more than 90 licenses and posts about $1.4 billion in annual revenue run rate.
What vertical AI means for adopting companies
For operating companies, vertical AI changes procurement from general tools to domain systems trained on relevant transactions, calls and workflows. Hippocratic AI builds voice agents for scheduling, pre-surgery preparation, discharge follow-up and chronic disease management without diagnosing or prescribing, and uses 31 models with 30 supervising one conversational model. Six investing health systems supplied 6 million real patient calls for fine-tuning, and the firm signed more than 60 enterprise clients including five large national payers. CloudWalk, with over 10 million active users and more than $2 billion in revenue, runs agents on hundreds of Nvidia Blackwell GPUs and automates 99% of customer support.
Adoption still requires checking data rights, safety architecture and integration paths rather than assuming a general model will suffice. Bloomreach reports almost half its customers use an AI agent, use of four agents grew 23-fold in a year, and third-party calls through Loomi Connect grow 83% month over month. Airwallex adds AI so customer agents can act across 80 to 100 economies, where interoperability and accuracy protect each transaction party. Harbinger builds trucks, RV chassis, storage and Army autonomous vehicles on shared lines and only adds businesses needing no new capital spending, showing the weight of manufacturing discipline.
The marker to watch is whether capital markets reward durability and scale in hardware-led AI during 2026. Bank of America executive JD Moriarty said AI and robotics firms now go public at far greater scale, public investors prize durable outsized growth, and activity leans toward hardware and semiconductors over software. CoreWeave was cited as a case where scale beat differentiation. If more tracked AI tailwind firms move toward listings on integrated capital solutions, vertical integration from GPUs to workflows will be confirmed as the business standard.
