I've heard a lot of conversations recently about how buying software is dead. Entire categories of software stock plummet on every Anthropic product announcement. So is the build vs buy in the enterprise paradigm over?
In my opinion, buy isn't dead, software companies have a strong future if they adapt and contrary to the general market vibe, I predict most of the tanking tech stocks will bounce back.
Where do existing software companies still have the edge?
Specialist knowledge, accountability and cost.
An LLM knowing about HR compliance is not the same as a vendor who has spent a decade encoding edge cases, jurisdiction-specific rules, and litigation-tested workflows into their product. They have captured this knowledge IP from year of pooling experience from many customers on a very specific problem. Broad ≠ deep and it's what LLM makers always say "seek professional guidance" or something like that. I know that LLMs can go surprisingly deep, but I just don't buy that they will replace highly specialist knowledge in highly dynamic environments. That gap is under appreciated in the "buy is dead" discourse.
Accountability and outsourcing the risk is 90% of the value of specialist SaaS software. Why would you take on a bunch HR, accounting, legal risk that those specialist platforms take off your shoulders. Companies are used to offloading this risk and they will factor this in to the build price. The old adage of "no one was ever fired for hiring McKinsey" will prevail here again.
Additionally, SaaS benefits from economies of scale and is often just cheaper than what it would take to run in house.
Opportunity for software vendors
Clearly software vendors need to adapt or they will get outcompeted though. The intrinsic value above needs to be delivered still and buyers will demand more customisability or simple want to interact with the core vendor IP through new mechanisms. If the legacy vendors don't seize this moment new players will eat their lunch.
I see a world where the legacy vendors package up their IP in LLM consumable chunks e.g. using MCP and deliver that as an option for you customers. The customer gets all the customisability and LLM interactivity for all their customisation needs as well as all off the benefits of working with accountable vendors with highly specialist knowledge.
But software companies have published public API's for years which could achieve exactly this concept of accessing a vendors core IP for customisation projects. The key difference here is that this has still required an in-house developers to stitch together and maintain. LLMs can now consume something like MCP immediately with almost zero fuss. There is a huge opportunity to become the go to LLM supplier of accounting, healthcare information, etc.
A nuance you could argue is that there's a difference between specialised software (payroll, healthcare records, financial compliance) and commodity software (basic project management, simple CRM, generic analytics) and on the surface the specialist knowledge edge seem to weaken, but, there is nuance with "commodity" SaaS. The current winners of those spaces won because of superior execution and better distribution rather then the core IP. But that is the exact same reason they can win in this new paradigm and having won their categories, it's has blessed them with a deep understanding of their customer's needs that's locked away from the LLM as organisational knowledge in internal playbooks, customer conversations, and years of product decisions that no foundation model has access to. This can be leveraged in the same way the more obvious specialist knowledge can.
So if software companies adapt quickly to the new reality, I think they still have a strong value proposition and can still be extremely relevant in the LLM era.