Anthropologic and OpenAI didn’t suddenly get interested in consulting. What they built forward deployed engineering organizations backed by billions from Blackstone, Goldman Sachs, TPG, and Advent, with service partnerships running through Bain and McKinsey, is not a consulting play. It’s a token economics play. And understanding the difference is the most important thing a mid-market leader can do right now.
Listen to the full episode:
Why the Trillion-Dollar Infrastructure Bet Depends on Your Organization Learning How to Use AI
Here’s the math that’s driving all of it. The AI infrastructure buildout is running at roughly a trillion dollars a year. Data centers, chips, energy, compute the capital being deployed is staggering. And it doesn’t get paid for by $20-per-month subscriptions. It doesn’t get paid for by advertising either. Google’s entire global advertising business generates around $165 billion annually. You can’t run ads to fund a trillion-dollar buildout. The only mechanism that works at that scale is token consumption. Organizations using AI deeply, at volume, across their workflows.
The problem is that most organizations don’t know how to do that. They buy the subscriptions. They run the pilots. They get excited in the demo. And then nothing changes at scale because nobody built the bridge between the tool and the workflow.
So Anthropic and OpenAI solved for that by funding embedded engineering organizations. Engineers who sit inside business functions, rethink how work gets done, and build AI-native workflows in place of the old ones. They’re not just implementing tools they’re driving the adoption that justifies the infrastructure investment. Forward deployed engineers are, in the most literal sense, the business model.
But those engineers, paired with McKinsey and Bain, are going to serve the Fortune 500. That’s where the money is concentrated. That’s where the contracts are large enough to make the economics work for a $4 billion deployment. And that leaves an enormous swath of mid-market and SMB organizations the ones that couldn’t have afforded McKinsey anyway with no viable path to the same kind of embedded AI expertise.
That’s the readiness gap. It’s not a knowledge problem or a motivation problem. It’s structural.
What’s Actually Happening Inside Mid-Market Companies Right Now
In the absence of embedded expertise, organizations have found their own version of forward deployment. They’ve hired smart, capable, motivated people often younger, often technically self-taught and pointed them at the problem. And those people are building. Fast.
What they’re building isn’t bad. A lot of it actually works. But there’s a pattern I keep seeing when I get called in, and it’s consistent enough to be almost predictable: the builders don’t know what they don’t know.
They’re not cutting corners out of laziness or bad faith. They’re moving fast because fast is what they were hired to do. But nobody has addressed the security vulnerabilities in what they’re building. Nobody has standardized the approach across projects. Nobody has thought through what happens when the first builder leaves and trains the second one and the second one inherits the assumptions along with the code.
As I put it on the podcast: a bad process is being institutionalized. The organization wakes up one day with a pile of functional tools, no CTO, no visibility into what’s running, and no idea what it would take to audit any of it. The apps work — until they don’t. And cleaning it up costs ten times what a governance framework would have cost at the start.
I spent 15 hours one weekend untangling a database architecture I had built myself over the course of several months, a project that a database architect would have taken weeks to fix in a traditional SQL environment. The tools helped. But the reason it needed fixing wasn’t technical. It was a process discipline problem. I had moved fast without keeping my back-end projects separated, and the mess compounded quietly until I couldn’t ignore it anymore.
That’s the version of this problem I get to solve in my own organization on a weekend. In a client organization with 50 employees and no internal CTO, it’s a different story.
The Governance Gap Is the Real AI Readiness Problem for Mid-Market Leaders
The forward deployed AI engineers mid-market companies actually need aren’t going to arrive from McKinsey. They’re going to have to be built either by developing the internal capacity or by working with external partners who can provide the combination of technical know-how, process expertise, and business context that makes adoption stick.
What I’ve found, particularly in mid-market organizations, is that three things need to be present for this to work. Someone technical enough to build the right solution. Someone with enough process experience to know how workflows actually function across departments. And someone who understands the specific business context the lingo, the data sensitivities, the stakeholder dynamics that determine whether a new tool gets used or quietly abandoned.
Uber’s pod model works brilliantly for Uber because their engineers have years of institutional knowledge. They already understand how the organization thinks. For most mid-market companies, that contextual knowledge has to be brought in from outside, which means the process expertise can’t live solely in the technical person.
The organizations getting this right are the ones building structured, repeatable approaches before scaling. Not waiting until they have a dozen apps running on different databases with no shared authentication and no audit trail. Not discovering their security posture after a vendor asks about it during due diligence. Getting the governance framework in place early, when it costs a weekend instead of months.
There’s also a more immediate decision most mid-market leaders are sitting on right now: the team-level AI platform question. When you have fifteen people each expending their own Claude or ChatGPT subscription, you have zero visibility into how the tools are being used, zero control over what data is being shared with them, and zero ability to build shared workflows that compound over time. The cost of upgrading to a team’s plan feels large until you price in what you’re losing by staying in the current arrangement.
And once you commit to a platform once your workflows, training materials, shared projects, and institutional knowledge are embedded in one ecosystem the switching cost becomes real. This is not a decision to make based on who has the better features this month. It’s a decision to make based on where your organization needs to be in 12 and 24 months, with eyes open about what the lock-in actually means.
What Mid-Market Companies Should Do Before the Gap Widens
The trillion-dollar infrastructure bet gets paid off by organizations that learn to use AI well, at scale, across their actual workflows. The forward deployed engineers being funded by Anthropic and OpenAI are going to accelerate that for the largest companies in the world. Mid-market companies are not in that conversation.
That’s the situation. And the response to it isn’t panic or a massive strategy overhaul. It’s getting honest about where your organization actually is.
Are your builders moving fast without a governance framework? That’s fixable but it’s easier to fix before the debt compounds than after.
Are you still running ten individual subscriptions with no centralized visibility? That’s a risk you can quantify and address.
Do you know what your AI workflows are actually doing, who owns them, and what breaks if the person who built them leaves? If the answer is no, that’s the conversation to start.
The organizations that will be in the best position in 24 months aren’t the ones with the most tools or the biggest AI budgets. They’re the ones that did the boring work of building repeatable processes, standardized governance, and genuine adoption before it became a crisis.
If you want to understand where your organization stands, the AI readiness assessment I built is a practical starting point: Explore AscendAI Launchpad
If what you’re dealing with sounds more complex than a self-assessment can address, I set aside time every week for exactly these conversations: Book a conversation with Kevin
