Unruggable Productivity

Are you giving Anthropic, OpenAI and Google an off switch on your business?

Hi friends,

This week we're getting into a theme we keep coming back to on the pod - unruggable productivity - and what it actually means for how SMEs should think about the AI tools they're building their business on.

Three news items worth your time as well, including some interesting numbers on what bringing AI in-house now actually costs.

The Free AI Audit we launched last week is still live at otherstuff.ai/ai-audit if you missed it.

Now to this week’s Good Stuff.

Are you giving Anthropic, OpenAI and Google an off switch on your business?

If you were building a business online between 2010 and 2015, Facebook Pages felt like a gift.

You posted, your followers saw it, you could build audiences and real commercial relationships relatively cheaply.

Then Facebook changed their algorithm and organic reach collapsed overnight.

Your still had all of your followers, the page was still yours, but now you had to pay to actually reach the audience you'd spent years building.

It was just a quiet policy change made in the interest of Facebook's actual customers, their investors.

This week on The Good Stuff, we explored the idea of unruggable productivity and why now might be the right time to ask yourself - who really controls my business?

Businesses are racing to hand their operational core - their workflows, processes, and institutional knowledge - to Anthropic, OpenAI, and Google.

In doing so, they are effectively giving those companies a kill switch on their own business.

If 90% of your business ends up running in Claude - is it still yours? Or is it theirs? Are you basically just like an affiliate for Anthropic at that point? And then they can just increase their costs. Are you going to migrate away from it? You probably can't, and they might not even let you.

Pete Winn - Episode 56 of The Good Stuff

That's the rug pull.

It's the same pattern as prior platform dependencies - AWS, Shopify, the App Store - except this time it's your company’s entire cognitive and operational infrastructure being handed over, not just your storefront.

The Kill Switch

When you hear "kill switch" you might think of someone literally turning off your software. But it doesn’t have to be that explicit to have a meaningful impact on your business.

The real kill switches are far more mundane, and just as dangerous, precisely because they don't announce themselves. They arrive as a pricing email, a terms update, a feature changelog. By the time you feel them, you're already too deep to leave.

Pricing power. Right now, Anthropic and OpenAI are in land-grab mode. Pricing is artificially low - they are paying for your dependency with subsidised compute. This is the oldest playbook in platform economics. AWS did it. The App Store did it. Google Workspace did it. The price you onboard at is not the price you'll pay in five years. The difference is that when AWS raised prices, you could migrate your database. When your operational brain is in Claude, what exactly are you migrating?

Terms changes. When your workflows, your customer data, your institutional processes live inside Anthropic's infrastructure, you are subject to their terms of service in perpetuity. They can restrict what industries they serve. They already do - there are categories of use they prohibit. Today that might not affect you. Tomorrow that could change and you will have no notice and no recourse.

Feature gating. The features you've built your workflows around today can be moved behind a higher pricing tier. Or deprecated. Or changed in ways that break your processes. You'll adapt - because you'll have to, but that cost of adaptation compounds every time.

Access withdrawal. Anthropic has already demonstrated willingness to cut off access to users and businesses that violate their policies - sometimes with minimal notice and contested reasoning. If your business runs on their infrastructure, you are one policy interpretation away from operational paralysis.

This Isn't Just Platform Risk. It's Something New

Businesses have always been told not to put all their eggs in one basket.

Don't be too dependent on one supplier, one customer, one distribution channel. That advice is old and sensible and completely inadequate for what's happening right now. Understanding why requires looking at what previous platform dependencies actually cost businesses - and what this one costs instead.

When companies built on AWS, what they handed over was their infrastructure. When merchants built on Shopify, what they handed over was their storefront. And when developers built on the App Store, what they handed over was their distribution.

In every case, the platform eventually extracted value - raised prices, changed terms, favoured their own competing products, and businesses were stuck because leaving was more painful than staying.

Those were serious dependencies, and businesses are still paying the price for them. But in each case, the business itself - the knowledge of how to operate, the judgment of how to make decisions, the institutional memory of what works - still lived with the people.

The platform had your storefront. It didn't own your brain.

With AI platforms, businesses are handing over the cognitive and operational core. When your customer service runs through Anthropic's infrastructure, when your internal knowledge base is encoded in their systems, when your project management and communications flow through their models, the dependency isn't just technical and commercial, it’s also cognitive.

Your processes have been absorbed into a system you don't own, and the longer that continues, the less they exist anywhere else.

What makes this harder to see clearly is that it isn't accidental.

VC-backed platforms call people users rather than customers because the actual customer is the venture capitalist, and what venture capitalists are buying is lock-in at scale.

This week, Anthropic announced an enterprise partnership with JP Morgan to drive enterprise adoption. It's a signal about who the real customer is and whose interests the product will be optimised for as it matures.

Small businesses are the users. JP Morgan is the customer.

The App Store pulled this on developers. Facebook pulled it on small business owners who built their audiences on Pages. This playbook isn’t new, but the depth of what's being extracted is.

So, What’s the Answer?

It doesn’t have to all be doom and gloom, but it’s important to ask ourselves - what does it mean to build a business that we actually own?

Your processes are documented in formats you control
Your agents run on infrastructure you host or can migrate
Your institutional knowledge is encoded in your own systems, not a vendor's
You have genuine optionality - you can switch models, switch providers, or bring it in-house
When AI pricing eventually normalises or commoditises, you benefit rather than being squeezed

The business that builds this way may be slower to get started and may have to accept more friction early on, but it’ll own something real at the end of it.

The business that races toward Claude and ChatGPT dependency will be faster to start but is ultimately building on rented ground.

This is part of the reason why we've been building our Wingman Suite.

It’s open-source agent infrastructure designed so that the intelligence remains yours. Your data, your workflows, your knowledge layer, all running on infrastructure you control.

We're still pre-launch, but if this is a problem you're thinking seriously about for your business, we'd like to hear from you.

We got into this and more this week in the Big Episode 56 of The Good Stuff

Three Things in AI for SMEs

1. Small businesses are now running an average of five AI tools — Margin Up

The SBE Council's 2026 Small Business Tech Use Survey, published last week, found that 82% of small business employers have invested in AI tools and the typical small business is now running a median of five different AI tools across the business, with most planning to add more over the next twelve months. Marketing and content creation are the top use cases, with workflow automation and customer engagement close behind. What it means for you: The interesting question has shifted from whether to use AI to whether your stack is actually integrated. Five disconnected tools is just five subscription invoices. SBE Council survey →Zenity

2. The economics of bringing AI in-house just shifted — Capital Up

Two technical guides published in the last few weeks have run the numbers on local AI servers for small business and the pay-back math has moved meaningfully. For a 10-person team with moderate usage, a $2,500 to $7,500 local server pays for itself in 3-5 months compared to cloud API costs. Setup time for a technically comfortable owner is now an afternoon rather than a project, with tools like Ollama exposing an OpenAI-compatible API in a single command. What it means for you: Owning your AI stack used to be a research project. The case is now closer to owning your server room twenty years ago. The compounding asset isn't the hardware, it's the workflows and data that stay on your side of the wall. ROI breakdown →Uscsinstitute

3. Inside Small Business is warning SMEs about a flooded AI tool market — Risk Down

A piece in Inside Small Business this week made the point that the explosion of new AI apps and platforms is creating a real due-diligence problem for owners. The advice from the experts quoted was that small business owners need to research customer reviews, security features, data handling, and privacy policies before handing over money, because the ease of building these tools means the market is now flooded with options that range from very good to actively dangerous. What it means for you: Cheap and easy to spin up cuts both ways. The AI tool you onboard this month is also a vendor you've quietly given access to your data, your workflow, and sometimes your customers. Worth treating with the same scrutiny as a payroll provider. Inside Small Business →PwC

That's all for today

If this resonated, we’d love for you to forward this to one person who might enjoy exploring these ideas too. See you next week!

Cheers,
Pete & Andy