I’m Glenn Gow, The Scaling Executive Coach and host of The Scaling Executive Podcast. In this episode, I sat down with Joshua Gould, Group CEO of thebigword, a global language services company operating in 80 countries with more than 15,000 contractors. When AI threatens to disrupt a core revenue stream, Gould holds that fact-based cost analysis, not panic, is the right response. He learned this in 1998, when a machine-translation tool threatened to erase 70% of his company’s revenue overnight. Actual client adoption came in under 1%.
That was the first of three AI disruptions thebigword has survived. Neural machine translation hit in 2013. Large language models hit in 2021. Gould’s company is still growing. His answer to why is not optimism. It is arithmetic.
This episode is for CEOs whose board members are pushing them to make fast, sweeping decisions about AI before the CEO has the cost data to justify them.
Key Takeaways
- Every major AI provider, including Anthropic, Google, and OpenAI, loses an estimated two to four dollars for every dollar of revenue it earns. This means AI pricing will rise, not fall, and CEOs who assume it will get cheaper are planning on bad data.
- thebigword cut its operating cost per revenue dollar from 20 cents to 5 cents, a 75% reduction, not by replacing humans with AI outright but by building API integrations and self-driving workflows around machine learning starting in 2013.
- AI interpretation reaches roughly 90% accuracy in specialized, high-stakes contexts like courtroom testimony and medical terminology. A 90% accurate witness in a murder trial is not acceptable, which is why thebigword still routes those cases to human interpreters.
- Gould treats his investors, including Susquehanna, to an unfiltered view of the business rather than a polished one, because a CEO who withholds bad news from the board loses the ability to get good counsel when a real threat, like AI, actually arrives.
- Gould projects the AI transformation of enterprise services as a 15 to 20 year shift, not a two to three year one, driven by rising chip costs and electricity demand that will outstrip power generation capacity for a decade or more.
Why AI Providers Losing Money Should Change How CEOs Budget for AI
Most CEOs are told that AI will get cheaper as the technology matures. Gould’s numbers say otherwise. The companies building the underlying models, Anthropic, Google’s Gemini, and OpenAI’s GPT, are not yet profitable on the AI they sell.
“The reality is my AI costs more than the humans because AI isn’t free,” Gould said. He estimates the major providers lose two to four dollars for every dollar of revenue they collect. That is not a temporary discount phase. It is a structural cost problem, and Gould expects it to get worse before it gets better.
Two forces are driving that trajectory. Chip prices have risen roughly 50% year over year, with no sign of leveling off. Electricity demand for AI infrastructure is set to outpace the world’s ability to build power stations for the next decade or two. When a CEO builds an AI budget on the assumption that costs will fall, Gould’s experience says that assumption is wrong.
This is why Gould pushes back on the idea that AI transformation is a two or three year event. “This is going to be a 15 to 20 year revolution. This isn’t a two to three year like a lot of people are saying. Those people are the ones that are trying to get investors for their own revolution,” he said. A CEO planning around a short, dramatic AI takeover is planning around someone else’s fundraising pitch, not the underlying economics.
Gould’s own company shows what a fact-based approach actually looks like. Rather than replacing his workforce with AI wholesale, thebigword built API integrations into clients’ content management systems starting after the 1998 near-miss, then layered in neural machine translation through a direct partnership with Google in 2013, then built self-driving workflows on top of that. The combined result cut thebigword’s operating cost from 20 cents per revenue dollar to around 5 cents, a 75% reduction that the company passed on to clients as lower pricing. That is a fact-based transformation built over more than a decade, not a panic-driven overhaul completed in a quarter.
Where AI Interpretation Still Fails and Human Expertise Is Still Required
AI has not replaced thebigword’s core interpretation service, and Gould is specific about why. In casual conversation, AI interpretation looks close to flawless. In specialized, high-stakes settings, the gap becomes a liability.
“It say goes down to about 90% accuracy, which is no good when you’re in a court setting and it’s a murder trial and the witness is only 90% accurate. I mean, that’s no good,” Gould said. Ninety percent accuracy sounds high until the setting is a courtroom, a hospital, or an immigration hearing, where a mistranslated medical term or legal phrase can change the outcome for a real person.
This is why the majority of thebigword’s business still runs on live human interpreters, dispatched on demand through a system Gould compares to Uber: a contractor turns up in person, on video, or on the phone within minutes. AI avatars now offer the same service, but Gould reserves them for lower-stakes interactions, not for court testimony or clinical care.
The distinction Gould draws is not AI versus human. It is a decision rule based on the cost of being wrong. When the accuracy threshold required by the situation exceeds what AI can reliably deliver, thebigword keeps the human in place regardless of what the technology can technically do.
Why Radical Candor With Investors Sharpens AI Decisions Under Pressure
Gould’s fact-based approach to AI decisions does not happen in isolation. It depends on a relationship with his investors that he describes as unusually direct for the private equity world.
“I actually treat my investors like I treated my father when he was the chairman of the company. Part shrink, part coach,” Gould said. Most CEOs, he explained, give their boards a polished version of the business. Gould gives Susquehanna, his primary investor since 2013, the real one, including the parts that are not going well.
That candor matters most when a board is anxious about AI and pushing for a fast, dramatic response. Gould works with a non-executive director, Jeff Kapinski, who regularly tells him when he has gone too far in a board meeting. “You can’t say that,” Kapinski will say. Gould’s response: “I’ve been saying it for the last three years.” That tension, a board member checking a CEO’s candor while the CEO keeps pushing the unfiltered version through, is what lets thebigword make cost-based AI decisions instead of headline-driven ones.
Gould believes this is the exception rather than the rule. “I’ll be the one and only company in their portfolio that will do that,” he said of his approach with Susquehanna. For CEOs facing board pressure to overhaul their business around AI on a shortened timeline, Gould’s experience suggests the fix is not a better AI strategy. It is a more honest relationship with the people asking for one.
The Framework: Making AI Decisions Without Panic
| Principle | What it means in practice | Named evidence from this interview |
| Fact-based cost modeling beats hype-driven urgency | Build AI capability incrementally, tied to measurable cost reduction, instead of announcing a sweeping AI transformation on a compressed timeline | thebigword cut operating cost from 20 cents to 5 cents per revenue dollar, a 75% reduction, through API integrations and machine learning workflows built between 1998 and 2013 |
| Human expertise stays mandatory above the accuracy threshold the situation requires | Set an explicit accuracy floor for each use case, and route anything below it to a human, regardless of AI’s overall capability | thebigword keeps human interpreters on court and healthcare cases because AI interpretation caps at roughly 90% accuracy in those specialized contexts |
| Radical candor with investors produces better crisis decisions | Give the board the unfiltered version of the business, including bad news, so the CEO has real counsel available when a genuine threat like AI arrives | Gould describes Susquehanna as the one investor in their portfolio receiving his fully unpolished view of the business, built over a five-year working relationship |
Quotes from This Episode
- “The reality is my AI costs more than the humans because AI isn’t free.” — Joshua Gould, Group CEO, thebigword
- “This is going to be a 15 to 20 year revolution. This isn’t a two to three year like a lot of people are saying. Those people are the ones that are trying to get investors for their own revolution.” — Joshua Gould, Group CEO, thebigword
- “It say goes down to about 90% accuracy, which is no good when you’re in a court setting and it’s a murder trial and the witness is only 90% accurate. I mean, that’s no good.” — Joshua Gould, Group CEO, thebigword
- “I actually treat my investors like I treated my father when he was the chairman of the company. Part shrink, part coach.” — Joshua Gould, Group CEO, thebigword
- “Thank you, Mr. Gould, for telling me, because we’re doing less than 1% of all our requirements.” — Honda executive, as recalled by Joshua Gould, Group CEO, thebigword
Frequently Asked Questions
Should CEOs panic when AI threatens to disrupt their core business model?
No. Joshua Gould, Group CEO of thebigword, has faced three separate AI disruptions to his business since 1998 and holds that fact-based cost analysis, not urgency, produces better outcomes. In 1998, a machine-translation tool threatened to eliminate 70% of his company’s revenue; actual client adoption came in under 1%, because the technology could not yet deliver what CEOs feared it could.
Why does AI cost more than human labor at enterprise scale?
Major AI providers, including Anthropic, Google, and OpenAI, are currently losing an estimated two to four dollars for every dollar of revenue they generate, according to Joshua Gould, Group CEO of thebigword. Combined with chip prices rising roughly 50% year over year and electricity demand expected to outpace power generation capacity for the next decade or two, Gould expects AI costs to keep rising rather than fall.
How should a CEO decide when human expertise, not AI, is still required?
Set an explicit accuracy threshold for each use case and route anything below it to a human. Joshua Gould, Group CEO of thebigword, applies this at his own company: AI interpretation reaches roughly 90% accuracy in specialized settings like courtroom testimony and healthcare terminology, which is not sufficient when the cost of an error is a wrongful verdict or a medical mistake, so those cases still go to live interpreters.
Executives Work with Glenn Gow to Scale Their Companies and Careers
Glenn Gow is The Scaling Executive Coach. He coaches ambitious executives into the CEO seat and CEOs into successful exits. With 25 years as a CEO and 5 years in venture capital, Glenn helps leaders scale their companies by scaling themselves first. If this conversation was useful, you can apply for executive coaching with Glenn Gow or apply to be a guest on The Scaling Executive Podcast.
