Every owner and CFO I talk to wants to know what AI can do for their business.

Almost none of them ask what AI needs from their business.

AI runs on data. Clean, connected, trustworthy data. In one place.

That is not a consultant’s opinion. Gartner predicts that through 2026, organizations will abandon 60% of AI projects that are not supported by AI-ready data. The models are getting better every month. The data underneath them is the part that decides whether they work.

Which is why a pitch I keep seeing lately deserves a closer look.

The “new” way to run finance

The ads and articles all describe roughly the same picture.

Start with a lightweight ledger. Add an AI tool for the close. Another for billing. Another for tax. Another for planning. Let agents connect the pieces and move work between them. Pick the best tool for every job and skip the big, heavy ERP.

It sounds modern. The products are well designed. Many of them are well funded. Some of them are genuinely good at the one thing they do.

But I have seen this before.

It used to be called best-of-breed

Not long ago, the standard advice for a growing company was to buy the best point solution for every function.

GL from one vendor. CRM from another. Inventory in a third. Billing somewhere else. Integrations holding it all together.

It worked. Until it didn’t.

Every system had its own version of the customer, the order, and the number. Month-end turned into a reconciliation project. Reports did not tie. Every upgrade on one side broke a connector on the other. And the real source of truth ended up being a spreadsheet someone kept up by hand.

That is a big reason unified platforms like NetSuite exist. Companies got tired of paying people to make their systems agree with each other.

The new way is the old way with better tools. The integrations are better. The problem of ten versions of the truth is not gone.

Why stitched-together systems break

Integrations are not a source of truth. They are translations.

Each tool defines things a little differently. What counts as a customer. When an order becomes revenue. Which entity owns a transaction. Every sync between two systems is a small decision about whose definition wins. Multiply that across eight or ten tools and you get a stack that is technically connected and practically inconsistent.

There are other costs that never make the demo.

Someone has to own each integration when it breaks. Someone has to manage users and permissions in every tool. Someone has to explain to the auditor how a transaction moved from one system to the next. And every vendor ships updates on its own schedule.

None of this is new. It is exactly the work finance teams spent the last twenty years trying to get out of.

AI makes the old problem more expensive

Here is the part that is different this time.

In the old best-of-breed world, people sat between the systems. A staff accountant noticed that the numbers did not match and went to find out why. Slow and expensive, but it caught things.

The new stack replaces those people with agents. Agents move faster. They also act on whatever data they can see.

Spread your data across ten tools and every agent works from a partial picture. The collections agent sees the invoice but not the open credit. The forecasting tool sees bookings but not the inventory constraint. Each one gives a fast, confident answer. Some of those answers are wrong, and nothing in the stack is positioned to catch it.

Fragmented data used to cost you time. With AI, it costs you decisions.

What AI actually needs from your business

Strip away the product names and the requirements are simple.

One set of master data. One customer record, one item record, one vendor record, used by every process.

Defined processes that live in the system. Approvals, routing, and handoffs that happen in the software, not in inboxes.

A complete audit trail. Every action, every approver, every change, captured in one place.

One set of permissions. So you know exactly what a person, or an agent, is allowed to see and do.

Get those four right and AI has something solid to work with. Skip them and AI will automate the confusion.

Why NetSuite is still the right foundation

This is where a unified platform earns its keep.

NetSuite runs finance, inventory, orders, purchasing, and customers on one database. The sales order, the fulfillment, the invoice, and the payment are all records in the same system, tied to the same customer. There is nothing to reconcile between them because there is no “between.”

Workflow and approvals are built into the platform. So is the audit trail. Permissions are managed in one place.

That is not a legacy idea. It is the foundation AI needs. And it is why Oracle is building AI directly into NetSuite instead of asking customers to bolt it on from the outside. The AI works across the same records your team already uses.

Where specialist tools still fit

This is not an argument against every tool that is not NetSuite.

Plenty of specialist tools add real value. Expense management, payments, tax calculation, and AP capture are good examples. Most NetSuite customers connect a few of them, and that is healthy. The best ones plug into NetSuite, write back to it, and treat it as the system of record.

That is the test. A good tool makes your core system more complete. A risky one asks you to make it smaller.

The question is not whether to use specialist tools. It is which system holds the truth.

Questions to ask before you buy the new stack

If you are outgrowing QuickBooks or an older system and the “new” way is on your shortlist, ask these before you sign anything.

  • Where does the customer master live, and what happens when two tools disagree?
  • How many integrations will we own, and who fixes them when they break?
  • Can we close the month without exporting data to a spreadsheet?
  • How do we show an auditor the full history of a single transaction?
  • When we add an entity, a currency, or a product line, how many systems have to change?

If the answers are vague, the old problem is still there. It is just better designed.

We wrote about the other half of this decision, the risk of migrating twice, in The New Wave of AI-Native Finance Tools Targeting NetSuite. And if you want to see what happens when AI is layered onto processes that were never defined, start with AI Fails Inside Companies That Never Automated the Basics.

The bottom line

The tools are new. The problem is not.

If AI is on your 2027 plan, the first line item probably is not AI.

It is the system underneath it.

If you are weighing that decision now, we are happy to walk through where your data lives today and what it would take to get it into one place. Book a strategy session.

Sources cited:

  • Gartner, “Lack of AI-Ready Data Puts AI Projects at Risk” (February 26, 2025)