TL;DR
- Applied AI discourse has collapsed into a doctrine about services and Forward Deployed Engineers. Adoption does not actually divide neatly into products and services.
- A third category is forming: malleable software, where the vendor ships a base product plus a harness and building blocks, and users extend it themselves.
- For small and mid-sized businesses with reasonably standalone systems, that is faster and cheaper than either custom development or conventional SaaS.
- Every software product will become extendable. The vendor stops being the pipe through which AI reaches the business and becomes the surfboard the user rides.
Much of the discourse in the business of Applied AI is clogged by the chant for AI Services religion and the high priests being ‘Forward Deployed Engineers.’ There is so much one hears about it; every frontier tech company and opportunistic finance bros peddling podcasts now have an AI services company of their own. The discourse has become a doctrine.
We started Moative to build and spin out ideas in regulated industries. Inevitably, we had to do consulting. We deployed smart folks to learn the language of the industries we serve before we could investigate ideas. It’s been two years. We have seen, first-hand, new forces that are taking shape, in how software gets deployed to enterprises.
Forward-deployed Consultants (a.k.a ‘Consultants’) are important to accelerate diffusion (a.k.a ‘Adoption). But adoption of applied AI software is not neatly divided into products and services.
There is a new category that is evolving: Malleable Software.
Software built by users
Todd Saunders, who is experimenting with enabling blue collar companies to build their own tools, did an informal pool on X. Here is what caught my attention:
What he calls flexible platforms is what I call as ‘Malleable Software.’
A few days ago, a company called Vendo launched. They speak to software product companies. Their pitch? Let your users extend your software.
We are launching our own CRM. Our pitch is that basic features come free and extending them is when you pay a one-time fee.
Why this beats custom development for mid-market
But I guess the more elegant version of this ‘extension’ of software is not custom-development. Custom development has its place in complex environments with a lot of legacy. For small and mid-sized businesses with reasonably stand-alone software systems, a base version with a harness (building foundation) and a set of lego blocks that the users themselves can use to extend the software is far more powerful, faster, and cheaper than custom development or even SaaS products. It also keeps the technology choices dull and well understood, which is the argument I made in Enterprise Grade is Dried Spit With A Budget.
Every software product will become extendable
Software product companies will rapidly enable harness capabilities and custom-builds on top of their platforms. It will be their answer to ‘vibe coding’ and ‘custom software development.’ This is the first good answer I have seen where I believe that software product companies can survive the AI wave, simply by being the surfboard that the users can use to take advantage of the wave.
Software companies will transform from being the pipe through which AI gets adopted by businesses to the surfboard that enables the users as they catch the right waves to surf.
Every software product company will provide a base software, a harness, a few pre-built blocks, a guardrail system of observability & security and put the roadmap of the software in the hands of its users. Consequently, software products are going to become complex. Every software company will become a platform by default. Mulit-tenant SaaS models will co-exist with multiple tenants. Self-hosted models will emerge too. We are experimenting with this approach. I set that out in The Dawn of Enterprise Agency Engineering.
The Ask
In the last few months we built a CRM, a project tracker, a HR performance management system, and an outbound email campaign automation system. If you are a small or mid-sized company that wants to extend these to fit your world, give me a shout.
We are giving away these products for free and extend them for you for a minimal fee.
Where Moative fits
Moative works as an AI operating partner with mid-market companies, which is why malleable software matters to us commercially rather than theoretically. The whole point of our model is that capability compounds inside the client rather than walking out with a contractor, and software your own people can extend is the clearest version of that. It is the same argument for keeping the data and the know-how on your side of the table, which I made in Don't shortsell your know-how to AI vendors.
Frequently asked questions
What is malleable software?
Malleable software is a product that ships as a base version plus a harness, a set of building blocks, and guardrails for observability and security, so that the customer's own users extend it rather than requesting features from the vendor or commissioning custom development.
How is it different from custom development?
Custom development builds a bespoke system, usually for a complex environment with significant legacy. Malleable software starts from a working product and lets users assemble extensions on top of it, which is faster and cheaper where the underlying systems are reasonably standalone.
How is it different from conventional SaaS?
In conventional SaaS the roadmap belongs to the vendor and every customer gets the same feature set. In malleable software the roadmap is partly in the customer's hands, because the extension mechanism is a product feature rather than a professional services engagement.
Why would software vendors give up roadmap control?
Because it is their most credible answer to customers building their own tools with AI. A vendor that cannot be extended gets replaced by something the customer builds. A vendor that can be extended becomes the platform that work happens on.
Who is this best suited to?
Small and mid-sized businesses with reasonably standalone systems and real process knowledge in-house. Large enterprises with heavy legacy integration are still likely to need genuine custom work.
What does this mean for software companies?
Every product becomes a platform by default. Products get more complex, multi-tenant and self-hosted models coexist, and the vendor's role shifts from delivering features to supplying a foundation other people build on.
