Two years into the AI boom, most companies still haven't seen the productivity gains they were promised. There's a reason for that, and it's not the technology.

Tim here, COO of Bridge AI Consulting, taking the pen from Dean this week.

As a bit of a history nerd, I like making sense of what's happening today by looking at how things played out the last time something this big came along. AI makes that hard. It's so powerful and transformational that most of the historical parallels people reach for fall apart pretty quickly. But factories adopting electricity in the early 1900s is the closest parallel to what's happening with AI right now. There's a clear lesson in how it played out, and the business owners who apply it are going to come out of this decade well ahead of the ones who don't.

This week:

• Why most companies are using AI the same way factory owners used electricity in 1905, and what to do instead
• A real client build: collapsing an entire pro basketball franchise's data sprawl into one connected database
• Two big OpenAI releases this week, and Google bets $40B that Anthropic still has more to give

Let's get into it.

When electricity arrived in factories in the late 1800s, the promise was a productivity revolution. For decades, it didn't show up.

Some factory owners tried to innovate. They ripped out the steam engine at the heart of their factory and dropped an electric one in its place. They expected transformation. They got modest savings instead. By 1900, almost two decades after electric motors became commercially available, they still accounted for less than 5% of factory power in America. The owners who had switched often felt cheated. The savings were real, but small. The promised revolution stayed promised.

The reason is that they had kept the rest of the factory exactly the same. Same giant central drive shaft running across the ceiling. Same leather belts cascading down to power every workstation. Same clusters of machines lined up around the shaft because they had to be physically close to it. New power source, old factory.

The unlock came when a different generation of factory owners stopped retrofitting old buildings and started designing new ones around electricity from the ground up. Every machine got its own small motor. Power could go anywhere. The floor plan was free. That's where the production line came from. That's where modular factories came from. That's where mass production came from. None of those things were possible while electricity was bolted onto the old structure. They required rebuilding the factory around electricity as the core.

That's where most companies are with AI right now.

Most companies are using AI to make their existing workflows a little more efficient. The marketer uses ChatGPT to draft an email faster. The analyst uses it to clean up a spreadsheet. The engineer uses it to autocomplete code. Same job, slightly less manual, modest gains. They feel like they're "using AI." Then they're underwhelmed when the productivity revolution doesn't arrive.

That's the steam engine swap. New power, old factory.

The actual unlock with AI isn't efficiency. It's that the work your business can produce is no longer capped by the number of people on your payroll. That's the constraint your business has lived under for its entire existence. And it's gone.

Once that lands, the work itself changes. Every employee gets agents tailored to their specific job, not just the analyst and the marketer. The marketer doesn't draft 5 emails a day, they ship 50 personalized ones. The analyst doesn't pull a quarterly report, they ship a live dashboard the whole staff can query. The recruiter doesn't skim resumes, they run every application against deep criteria.

The human becomes the architect, the verifier, and the shipper. The AI does the manual work, and a lot more of it than a human ever could.

One shared chatbot pointed at one shared data source isn't an AI strategy. It's a more efficient steam engine.

Here's the place the metaphor breaks down. Electricity took 40 years to deliver on its promise. AI is moving on a different timeline. The companies that figure this out in the next two years are going to compound advantages that latecomers spend the rest of the decade chasing. The window to be early is small and closing fast.

Companies don't need an AI strategy. They need to rethink their work around a constraint that just disappeared. Not swap. Not plug in. Rethink and rebuild.

That's harder than buying Claude licenses. It's also where the actual money is.

If you want to talk about what rethinking the work actually looks like for your business, not in theory but in your actual workflows with your team, reply to this email. That's what we do at Bridge AI.

The Flying Cows are a professional basketball team. Like a lot of pro sports operations, they had data everywhere. Tickets in one system, fan info in another, sponsors in a third. Players, games, and groups had their own corners too. Some of it was duplicated across multiple platforms. Some of it was out of date.

If an owner asked a basic question, like "who came to the last three home games and also bought concessions?", someone in the front office had to pull from four different platforms, stitch it together in a spreadsheet, and hand back an answer that was already stale by the time it landed. It took hours. And nobody fully trusted the result.

We built one connected database that holds tickets, fans, players, games, sponsors, and groups, with every record linking to the others. Ticket data syncs in nightly, completely autonomously. The whole thing is wired into Claude, so anyone in the front office can ask questions in plain English, generate reports, build dashboards, and update records without touching the underlying systems.

Pulling a fan or group subset that used to take hours now takes a couple minutes. Dashboards refresh automatically. When a number changes anywhere upstream, the answer changes everywhere downstream. The owners and front office get the answers they need on demand, and they can trust them.

A human still spot-checks records before any outreach goes out. The database does the assembly. The team does the judgment.

This is the deep dive in miniature. We didn't make the old workflow faster. We replaced it.

OpenAI ships GPT-5.5, six weeks after the last one

OpenAI released GPT-5.5 on Thursday, its second major model upgrade in six weeks. It's better at agentic coding, deep research, computer use, and "moving across tools until a task is finished." Available to Plus, Pro, Business, and Enterprise users across ChatGPT, Codex, and the API.

My take: Six weeks. That's the gap between GPT-5.4 and GPT-5.5. A year ago, a major model release was an event you'd circle on the calendar. Now they're shipping faster than most companies can update their internal training docs, and the improvements between releases are getting bigger, not smaller. The frontier is moving exponentially. If you tested an AI tool last quarter and decided it couldn't do the job for your business, test it again. It can probably do the job now.

READ THE TECHCRUNCH REPORT →

ChatGPT launches Workspace Agents, its enterprise answer to Claude

OpenAI also dropped Workspace Agents this week. They're shared, always-on Codex agents that plug into Slack, Salesforce, and 60+ other enterprise apps. They run in the cloud. Teams build them once, share them across the org, and improve them over time. Free during research preview through May 6, then credit-based pricing.

My take: This is OpenAI catching up to Claude on the enterprise side. Persistent agents, deep tool integrations, shared workflows. These were Claude's differentiators six months ago, and we'd have said it wasn't even close. Now Workspace Agents puts ChatGPT in the same conversation. We still build on Claude for our client work and we still think it's the right call. But the gap is closing fast, and any business owner picking an AI vendor right now should know that committing to one of them long-term is a bet, not a foregone conclusion.

READ THE VENTUREBEAT REPORT →

Google to invest up to $40B in Anthropic

Google announced plans to invest up to $40 billion in Anthropic this week. $10B now, up to $30B more tied to performance milestones, at a $380B valuation. Amazon is also adding another $5B in a separate deal. Between the two hyperscalers, well over $50B is now flowing into Anthropic.

My take: The story for the last few years has been that OpenAI had the capital firepower and Anthropic had the better product. As of this week, Anthropic has both. Google and Amazon have collectively put north of $50B into the same company, which gives Anthropic comparable resources, comparable infrastructure, and comparable runway to OpenAI. Somewhere in the next 18 months, one of these two is going to be the first frontier AI lab to file for an IPO. That race is going to define the next phase of this industry.

READ THE CNBC REPORT →

OpenAI also launched Images 2.0 this week, and it's the new best image generator on the market. Native reasoning, 2K output, multi-image consistency, and (finally) text in images that actually renders correctly. If your team produces marketing visuals, social graphics, or pitch deck imagery, this is the one to be using.

READ THE OPENAI REPORT →

That's it for this week.

If you found this useful, forward it to someone who's been kicking the tires on AI for their business and wondering why it isn't doing more for them. They probably need to hear the steam engine comparison.

And if you want to talk about what AI could actually do for your business, not in theory, but in your workflows, with your team, on your timeline, just reply to this email. That's what we're here for.