AI Guidance for Physical Work

Workers in field service, manufacturing, and healthcare need months of training before they're productive. Real-time AI coaching through phones, AirPods, and smart glasses can make anyone effective immediately—'turn off that valve,' 'use the ⅜ inch wrench.'

Training a new HVAC technician takes 6-24 months before they can work independently. But what if they could be productive on day one—not because they know everything, but because an AI expert is in their ear, seeing what they see, guiding every step?

This isn't about replacing training. It's about making every worker as effective as your best worker, immediately.

The Technological Convergence

Three capabilities have matured simultaneously:

1. Multimodal AI Can Now "See" Work GPT-4V, Gemini, and Claude can identify equipment from photos, read model numbers, interpret error codes, and understand spatial relationships. They can look at a circuit breaker panel and tell you which breaker controls which circuit.

2. Hardware Is Already Everywhere

• •97% of field workers carry smartphones with cameras
• •AirPods/wireless earbuds are ubiquitous—voice guidance without looking at a screen
• •AR glasses (Meta, RayBan, Xreal) are reaching consumer price points
• •Edge AI chips enable on-device processing for low-latency guidance

3. Labor Economics Make This Viable With median electrician wages at $62,350 and experienced HVAC techs earning $150,000+ in major markets, even modest productivity improvements justify significant software spend. A 20% reduction in callbacks alone could save $15K+ per technician annually.

Real-World Scenarios

Scenario 1: The First-Day HVAC Tech Marcus just finished trade school and starts his first job at a commercial HVAC company. Traditionally, he'd shadow a senior tech for 3-6 months before taking calls alone. With AI guidance:

Day 1: He arrives at a service call. He opens the app, points his phone at the rooftop unit. The AI identifies it: "Carrier 48TM028, installed 2019. Common failure modes for this unit are compressor capacitors and contactor issues."

Marcus: "It's making a clicking sound."

AI: "That clicking typically indicates the contactor is trying to engage but failing. Look at the electrical panel on the left side—I'll walk you through testing the contactor."

15 minutes later, Marcus has diagnosed a $40 part replacement that would have cost the customer $800 if it had been misdiagnosed as a compressor failure.

Scenario 2: The Rural Nurse Sarah is a new graduate nurse at a rural clinic in Wyoming, 90 miles from the nearest hospital. She encounters a patient with an unusual skin presentation. She snaps a photo (with consent) and the AI provides differential diagnosis support, suggests questions to ask, and recommends when to escalate. This doesn't replace the physician—it helps Sarah collect better information before the telemedicine consult.

Scenario 3: The Manufacturing Line Toyota's Georgetown, Kentucky plant has 8,000 workers. When a new model launches, thousands of workers need to learn new assembly procedures. Currently, this involves weeks of training and months of reduced productivity. With AI guidance overlaid via tablet screens at each station, workers can follow procedures precisely while the system flags deviations in real-time.

The Business Models

Model 1: B2B SaaS to Service Companies Sell to HVAC, plumbing, and electrical companies. Pricing: $200-500/month per technician. Value proposition: reduce training time, decrease callbacks, improve first-time fix rates.

ROI math for a 50-tech HVAC company:

• •Current callback rate: 15%
• •With AI guidance: 8%
• •Savings: 7% × 50 techs × 5 calls/day × 250 days × $150/callback = $131,250/year
• •Software cost: 50 × $300/month × 12 = $180,000/year
• •Breakeven: Need 10% callback reduction

Model 2: Vertical Integration (The Full-Stack Approach) Don't sell software to HVAC companies—become an HVAC company. Hire people with basic mechanical aptitude, train them in 4 weeks instead of 6 months, and use AI guidance to maintain quality. This is the "Amazon approach"—use technology to do something that was previously impossible, not just incrementally better.

Potential advantage: You capture the entire margin, not just software fees. You can price 20% below competitors while maintaining higher quality.

Model 3: Platform for Independent Workers Enable anyone to become a "skilled" tradesperson. The plumber's version of Uber. Worker gets the calls, AI provides the expertise, platform takes a cut. This democratizes access to skilled trade income for people who couldn't afford 4-year apprenticeships.

What's Already Being Built

• •XOi Technologies: Field service video + AI for HVAC/plumbing (raised $66M)
• •Augmentir: AI-powered connected worker platform for manufacturing
• •Parsable: Digital work instructions for frontline workers
• •SightCall: Visual assistance platform for field service
• •Help Lightning: AR-based remote expert guidance

Most of these focus on connecting remote experts to field workers, or providing digital work instructions. The opportunity is fully autonomous AI guidance—no remote expert required.

Technical Challenges to Solve

1. •Latency: Voice guidance needs <300ms response time to feel natural
2. •Context persistence: AI needs to remember what was done 5 minutes ago
3. •Safety: AI must know when NOT to give advice and when to escalate
4. •Equipment recognition: Need to identify equipment from partial views, worn labels, unusual angles
5. •Procedure accuracy: A wrong instruction could cause injury or property damage

Market Sizing

| Segment | Market Size | Rationale | |---------|-------------|-----------| | Field service (HVAC, plumbing, electrical) | $50B+ | ServiceTitan alone does $500M ARR serving this market | | Manufacturing frontline workers | $30B+ | 12M manufacturing workers in US | | Healthcare clinical support | $20B+ | 4M nurses, 2M home health aides | | Construction | $15B+ | 8M construction workers |

Even capturing 1% of field service training and productivity spend is a $500M opportunity.

What We're Looking For

The winning company will:

1. •Start with a wedge: Pick one trade (HVAC is ideal—complex enough to need help, standardized enough for AI to learn)
2. •Build proprietary data: Every interaction trains the model. First-mover advantage compounds.
3. •Solve the hardware UX: Phone-in-hand is awkward. Voice + AR glasses could be transformative.
4. •Navigate licensing: Work with trade boards, not against them. AI guidance doesn't replace licensing requirements.
5. •Prove safety: Document that AI-guided workers have equal or better safety records than traditionally trained workers.

The company that makes a day-one worker as effective as a five-year veteran will transform the skilled trades economy.