Show Notes
A CEO told Justin Watt his company was ready for AI. "We've got our data architecture together," he said, giddy. Justin asked to see it. The guy pulled up an Excel file. The filename? Data Lake.
Funding stage: Bootstrapped. Switchboard is a two-year-old, referral-grown, 20-person consultancy with no disclosed outside funding, deliberately scaling slowly on client revenue. Textbook bootstrap.
That moment is the whole episode in miniature. Justin Watt, co-founder of Switchboard, studied psychology, not computer science, and that turns out to be his unfair advantage. After stints at IBM and MetaLab, where his teams built products for Uber and Amazon and helped design Slack, Justin realized the hardest part of every technology project is never the technology. It is the humans. Every business challenge is a human challenge wearing a software costume.
Switchboard works with mid-market companies, the $50 million to $500 million crowd, the businesses old enough to have 40 years of legacy process and young enough to actually change. These companies think they are AI-enabled because they bought everyone a Claude license. Meanwhile, month-end close runs through one person's spreadsheet that nobody else can read, and if that person quits, the business forgets how it works.
Justin's fix is unglamorous and devastatingly effective: map the real workflow, not the org-chart version. Find where humans are doing machine work. Inject AI at the steps where it actually moves the needle. Keep humans in the loop everywhere else. The result is not layoffs, it is smart people finally doing smart work. In Justin's experience, less than 5% of leadership conversations are about cutting headcount. The conversation is always about the endless pile of work standing between the company and its goals.
Along the way, Ryan and Justin cover the AI washing epidemic, blaming layoffs on AI to cover up old hiring mistakes, why frontier lab doom marketing blew up in everyone's faces, the death of "bring your whole self to work," quiet quitting as cowardice, Garth Brooks selling his catalog for a rumored $2 billion, ravens that speak English, and the most surreal government website in existence.
Named Frameworks
The Factory Line Injection Model
Justin's mental model for where AI actually belongs in a process.
- →Picture any business process as a 30-step assembly line.
- →AI does not apply to all 30 steps, and pretending it does is how pilots die.
- →It transforms steps 2, 7, and 12 while humans run the rest.
- →Map first, inject second, and reclaim up to 30% of your team's week.
Single Player vs Multiplayer AI
Why a license for everyone is not a strategy.
- →Dropping a spreadsheet into a chat window is single player mode, and it works fine for one person.
- →A business is a multiplayer environment: 20 brokers, a CRM with half the data, spreadsheets with a third, and a phone call nobody logged.
- →Multiplayer AI requires a source of truth with an API, not a file 50 people email around.
- →If your AI strategy is 'everyone got a license,' you are playing single player in a multiplayer game.
The Real Spreadsheet Audit
Find the hidden process that actually runs the company.
- →Every company has an official process and a real one.
- →The real one usually lives in a spreadsheet built by one person who quietly runs month-end.
- →Find the real spreadsheet, document the hidden logic, and remove the single point of failure.
- →Bonus test: if a key person left tomorrow, could anyone close the books?
Strategy as Forward Deployment
Switchboard's anti-slide-deck approach to strategy.
- →Justin is anti management consulting in the slide-deck sense.
- →Switchboard's strategy phase means sitting with the people in the weeds for two to eight weeks.
- →Output: a roadmap of real opportunities ranked by needle movement, not spaghetti pilots.
- →Then a monthly retainer builds the roadmap, because Rome was not built in a sprint.
The Meritocracy of Being Wrong
How Justin keeps the power dynamic from killing good ideas.
- →Justin tells his team he loves to be told he's wrong.
- →The power dynamic breaks when the leader always has to win.
- →Yin and yang co-founding: Justin and Jo divide by genuine preference, so lane disputes never happen.
Founder Experiment to Run This Week
The Bus Test Audit. Pick your most critical recurring process, such as month-end close, client onboarding, or quoting. Ask one question: if the person who runs this disappeared for 30 days, could we still do it? If the answer involves a spreadsheet only one human understands, you found your first AI project. Map every step of that process, mark which steps are humans doing machine work, and identify your steps 2, 7, and 12. Do not buy a single license until the map exists.
Glossary
Mid-market
Companies doing roughly $50 million to $500 million in annual revenue.
AI washing
Publicly attributing layoffs or changes to AI to mask ordinary mismanagement.
Data lake
A centralized repository for structured and unstructured data. Not an Excel file, no matter what the filename says.
Source of truth
The single authoritative system where data lives, ideally with an API.
API
Application Programming Interface, the connection point that lets tools talk to each other.
Forward deployed engineer
A technical operator embedded inside the client's actual workflow rather than advising from a deck.
Change management
The human side of rollout, getting people to rethink how they work.
EOS / Visionary-Integrator
The Traction framework pairing a big-picture founder with an operations-focused counterpart.
Quiet quitting
Doing the minimum while staying on payroll.
LLM
Large language model, the engine behind tools like Claude.
Q&A: What Founders Ask After This Episode
What does AI transformation look like for a mid-market company?
Map the actual workflow, find where humans do machine work, and inject AI at specific steps while keeping humans in the loop, rather than buying licenses and declaring victory.
Why do most corporate AI rollouts fail?
Companies apply single player tools to multiplayer environments and point AI at fragmented data spread across spreadsheets, CRMs, and unlogged conversations.
Is AI actually replacing jobs in 2026?
In Justin Watt's client conversations, less than 5% involve layoffs. The goal is reclaiming time for higher-value work, and many AI layoff claims are PR cover for past hiring mistakes.
How much does an AI strategy engagement cost?
Switchboard's strategy projects run two to eight weeks at $20,000 to $100,000 depending on complexity, then convert to a monthly build retainer.
What industries benefit most from AI consulting?
Professional services, insurance brokerages, and legacy mid-market companies with operational maturity but outdated, person-dependent processes.
Five Founder Questions This Episode Answers
- →How do I know if my company is actually AI-enabled or just paying for licenses nobody uses?
- →Where exactly in my operations should I deploy AI first to get the biggest time savings?
- →How do I de-risk the business from the one employee whose spreadsheet secretly runs everything?
- →What should an AI strategy engagement cost, and what should I demand it includes?
- →How do I lead an AI rollout without triggering fear, resistance, and burnout in my team?
URLs Mentioned in the Episode
- Switchboardhttps://withswitchboard.com
- Justin Watt on LinkedInhttps://www.linkedin.com/in/wattjustin/
- AI for Foundershttps://aiforfounders.co
- Inbox Alchemyhttps://inboxalchemy.co
- SPCA Internationalhttps://spcai.org
- war.gov/ufohttps://www.war.gov/ufo
- Sunohttps://suno.com
