What a demo does not show
Six months after signing the contract, we discovered that half the features nobody asked for are not being used. The lawyers find the interface very busy. Customer service takes a week to reply. And the proposal to make the platform actually work the way we need it to work arrived with a fifty thousand price tag attached.
Does this sound familiar? None of this is anyone's fault in particular. If you need a car, you probably need something reliable that gets you places. You do not need your new car to blend you smoothies. And we all have examples of vendors behaving like you have a smoothie problem that only their car can solve. The platform sold was never really a product designed to solve a real need. It was advertised as an "ecosystem" — a portfolio in which one or two components may or may not solve a real problem the firm has, and the rest are sold on the strength of the brand.
So how do firms avoid walking into this?
The maths matters before the logos do
Numbers help here. SaaS Capital, an independent research firm that has tracked the sector for over a decade, currently puts the median public software business at roughly six to seven times annual revenue. M&A advisors publishing comparable benchmarks put the long-run median for software acquisitions lower still, in the three-to-five times range. AI-specific companies command a premium, but even so, recent sector analysis puts the median fundraising multiple for AI businesses at around twenty-five to thirty times revenue. A healthy market produces those benchmarks.
Against them, Bloomberg Law reported last autumn that at least one prominent legal AI vendor was valued at roughly fifty times its annual recurring revenue. That number is not an accident and it is not really a statement about current cashflow. It is a bet by sophisticated investors that the company will grow into the price. For the bet to pay off, the product has to become extraordinarily sticky — sticky enough that when the price rises, or the direction shifts, or the support quality slips, the customer still does not leave. Most products carrying that kind of multiple are not yet that sticky, and the gap between what the valuation demands and what the product currently delivers is where the buyer's risk quietly sits.
The Harvard Law School Forum on Corporate Governance has summarised from the academic research: venture-backed companies fail at a rate of roughly seventy-five percent. Three in four of the overhyped names in legal technology today will not be around in five years, at least not in their present form.
Vendors running on these numbers are behaving the way their capital structure requires them to behave. The mistake is on the buyer's side — assuming a vendor in this position can afford to be a neutral infrastructure partner, when the maths says they cannot.
Product stickiness and lock-in
It is tempting to use the phrase "vendor lock-in risk", but let's try to dig a little deeper. Complex business software does not behave like a TV streaming subscription. Buying a serious platform involves months of project management, preparation, role-based testing, integration work, template building and the kind of internal change management that costs both sides real money. A good vendor understands that, and this investment is done to produce a customer who stays for years because the value is real.
What happens if the product or service does not deliver on promises? How much appetite a firm should have for trying new things depends heavily on its size. A twenty-person firm is in a genuinely good position to experiment. A wrong choice costs a few months of subscription and some inconvenience. Small firms should be trying things, and, honestly, they should be trying more than they currently are.
A two-hundred-person firm is playing a different game. The legal industry is famously slow and conservative. Some firms were happy with WordPerfect until the moment a major client was not. Technology change in professional services industry is rarely driven by internal curiosity. It is driven by client pressure, and once a major client starts asking questions during the panel review, a technology question becomes a reputation question.
None of that means larger firms cannot experiment. It means experiments need to be ring-fenced. One good example of this approach would be running pilots inside a single practice group rather than rolling new tools out across the whole firm, and treating the pilot as a genuine test with a real decision at the end. That approach lets the firm learn something about the vendor without too much risk at stake, and it produces the kind of evidence that makes the eventual firm-wide decision defensible.
A vendor as a partner
Technology procurement in professional services is often treated as an IT job. That is not wrong — IT teams handle data security, integration, and they keep the firm out of trouble. Those are the foundations, and no technology choice can succeed without them.
But technology decisions are now strategic decisions — whether the vendor will still exist in its current form, whether their strategic direction will still match the firm's, whether the people who sold the platform will still be the people running it. Larger firms have started answering this with structure by creating innovation departments. These teams work closely with business, marketing and IT, with a goal to think about how technology choices shape the firm's commercial future. The titles vary, but the function is the same: someone senior who puts those questions in front of the partners and is able to articulate them.
How do you tell the difference between a vendor worth your time and a supplier with a marketing budget? Put them in a room with your partners and ask the uncomfortable strategic questions. The ones who answer without flinching are the partners you are looking for. The rest have already told you everything you need to know.