If We Have AI, Why Do We Need Anything Else?

This week I heard the same question from two different people. One was an in-house lawyer. The other was an industry expert I count as a mentor. Both described the same thing: companies are spending heavily on AI, and teams feel real pressure to put it to work, often with very little guidance on how.

That money has to come from somewhere, and much of it is coming out of other technology spend. So when legal, or finance, or HR asks for a tool of its own, the question lands fast: if we have AI, why do we need anything else?

It is a fair question. It is also not just a legal one.

Start with the system nobody would switch off

Picture the CFO in that same budget meeting. Would anyone suggest shutting off the ERP or the accounting system now that the company has AI?

No one would. The reason is worth saying out loud. AI can read the ledger, explain a variance and draft the memo on why margins moved. It does not keep the ledger. It does not run the month-end close, hold the chart of accounts, enforce who can approve a payment, or keep the audit trail your auditors will ask for. Those jobs belong to a system built for how finance works.

AI makes that system far more useful. It does not replace it.

What AI buys you, and what it doesn't

When a company invests in AI, it is buying intelligence: a model that can read, write, reason and answer questions in plain language. That is a real investment and worth making.

What it is not buying is the operating system for the work:

• The workflow. What happens first, what happens next, and who owns each step.
• The record. The same facts captured the same way every time, so you can count them later.
• The rules. Who can see what, who approves what, and what has to happen before something closes.
• The communication. Who gets told what, and when, without anyone having to chase it.
• The history. What was decided, by whom and why, still there a year from now.

Intelligence answers the question in front of it. It can even remember the conversation. What it does not do is structure the data, run the workflow, get the right work to the right person at the right time, or keep the record secure and retained for as long as you need it. That is the system's job.

Where the money is moving

I have written before that software should cost less now. That does not mean technology budgets shrink. Part of that spend is moving to AI models, and that is healthy.

But it changes what you should expect from the software you keep. It should have AI built in. And it should be able to work with the AI your company has already paid for, so you are not buying the same intelligence twice.

A few ways to picture it

The engine and the car.

A powerful engine sitting on the garage floor goes nowhere. It needs a frame, steering, brakes and a dashboard. Nobody argues about which one matters more. You need both to get anywhere.

The brilliant new hire.

Hire the smartest person you can find and give them no process, no filing system and no idea who to hand work to. They will answer every question you ask. The work still falls through the cracks, because smart and organized are two different things.

Electricity and the factory.

When factories first got electric motors, many simply swapped them in where the steam engine had been, and not much changed. The big gains came later, once factories were redesigned around the new power: how work moved across the floor, who did what, and in what order. The power was never the whole answer. The way the work was organized decided what it was worth.

That last one is where a lot of companies are with AI today. The power has arrived. The work has not been reorganized around it yet.

How AI and your systems make each other better

This is not a choice between AI and systems. They feed each other.

• The system makes the AI smarter. Give AI clean, structured records with the right permissions and its answers get better. Point it at scattered email threads and chat logs and the best it can do is guess.
• The AI makes the system easier. Nobody has to fill out a long form or learn a new screen. People ask in plain language, where they already work. The AI sorts the request, fills in the fields and writes the summary.
• The loop keeps paying off. Every answer the system captures is one the AI can reuse, so a question only has to be answered once. Every record the AI helps create makes next quarter's report possible. The more you use one, the better the other gets.

What is bring your own model, and why is it next?

Most software with AI today runs on the vendor's own models. That makes sense for now. But the next step is already visible. A company that has invested in its own AI platform will want the software it buys to run on that platform, not on a second model it has to vet, secure and pay for.

That is a big shift for software, and it asks more of the people who build it. To me it is the real test of being AI native. If the model is designed to be swapped, pointing to a different one should be easy when the customer asks.

It does raise the bar on trust. The vendor still has to be clear about what the model is doing and show its work, with as much traceability as it can offer. That is harder when the model belongs to the customer. I expect it gets easier over time.

This is not a roadmap slide for us. We designed LegalOperator from day one so the model can be swapped, not hard-wired. That groundwork is already in place, and it is how we will support customers who want to bring their own.

What this looks like for an in-house legal team

Legal is where I spend my time, so here is the version I know best. An employee asks a question in chat. AI answers from the company's own policies. If it can't, or shouldn't, the question goes to the right attorney with the whole conversation attached. Real legal work becomes a tracked matter. If it goes to an outside firm, the invoice gets checked against what was agreed. Every step lands in the same record.

That is what we built LegalOperator to be: the AI-native operating system for in-house legal teams. Intake, matters and outside counsel spend live in one record, with AI as the intelligence running behind it.

And it already works with the AI your company bought. Through our MCP Connector, tools like Claude or ChatGPT can work directly with LegalOperator, so the investment you have made keeps working for legal too.

Before you decide AI covers it

Five questions worth asking, whatever the department:

1. Where does the record live when the chat window closes?
2. Is every request captured the same way, so you can count it next quarter?
3. Who approved it, and can you show it?
4. Who gets told what happens next, without having to chase?
5. Can you answer the board's question with numbers, not a guess?

If the answers are "nowhere," "no" and "nobody," the AI is not the gap. The operating system is.

Common questions

The short answer

If someone asks why you need anything besides AI, here is mine. Smart and organized are two different things. AI is the smart. The system is the organized. Buying one does not replace the other, any more than AI replaced your accounting system. Put them together and each one makes the other better.

Craig Raeburn

Craig Raeburn is a cofounder of LegalOperator, where he leads go-to-market. He's spent 20+ years in B2B software, often brought in at the messy inflection points: turnarounds, acquisitions, and integrations.