The ERP implementation problem
The pitch for an ERP system is easy to like: one system for the whole business, with inventory, orders, finance and operations in one place. The part the pitch leaves out is the project that comes before anyone can use it.
That project is the reason many small and mid-size businesses never buy one. They stay on spreadsheets and a handful of tools that do not talk to each other, and they pay for it in retyping and mistakes instead.
Why traditional implementations take so long
Traditional ERP software is built to be configured. It ships with a large set of options that try to cover every kind of business, and somebody has to work out which of those options match yours. The usual project runs through the same phases:
- Discovery and scoping: workshops to document how the business runs today
- Configuration: consultants mapping your process onto the structures the software already has
- Data migration: exporting, cleaning and importing old records
- Testing: finding the gaps between what was configured and what people need
- Training: teaching the team a new way of working
- Go-live: the stretch where everything is real and mistakes have consequences
Each phase depends on the one before it. A delay early on pushes everything back, and the further in you are, the more it costs to change your mind.
The cost is more than the invoice
Licenses and consulting fees are the visible cost. The less visible one is attention. The people who know how your business runs are the same people the project needs in workshops and testing sessions. The work lands on your busiest staff at the moment you are also asking them to change how they work.
There is also the cost of fit. When changing the system is expensive, most businesses end up bending their process to match the software.
What changes when you can describe the software
AI builders start from the other end. Instead of configuring a finished product, you describe how your business works and the software is built around that description.
In Unstack that looks like this. You tell the AI builder: "I need to track customer orders from quote to delivery, and I want an alert when an order sits untouched for two days." The builder creates the tables that hold the data and writes the pages your team will use. On plans that include automations, it sets up the alert too. You look at the result, say what is wrong, and the builder changes it.
The step that used to need a specialist, turning a description of the business into working screens and data, is the step the AI now does.
From "implement, then customize" to "describe, then adjust"
In a traditional rollout, customization comes last and costs extra. Here it is the starting point, and that changes a few things:
- You start with something real. You react to a working page instead of a requirements document.
- Changes are small. When a process changes, you describe the change. There is no change request to file.
- The system fits the business. The tables and pages are named after your work, not after a generic setup.
- You do not have to start from zero. The Marketplace has templates, such as a contacts CRM or invoicing, that you install and then change.
What you get in Unstack today
- Tables and pages. Your data lives in tables you define. Pages are real code that the builder writes and builds, and you can open the code if you want to.
- Automations. Workflows that run when a record changes, on a schedule or when another system calls them.
- A customer portal. Customers sign in and see their own records.
- Payments. Invoicing and a point of sale, processed through Stripe.
- Integrations. Connections to thousands of other apps, including QuickBooks and Xero for accounting.
- AI agents. AI employees you set up with instructions, tools and triggers.
- Roles. Control over which pages, tools and data each person can reach.
Pricing is one flat price for the workspace with unlimited users, and there is a Free plan. Which features you get depends on the plan. The pricing page lists what each one includes.
What AI does not do for you
It does not decide how your business should run. If your team has not agreed on how orders get approved, no software will settle that for you.
It does not clean your data either. Old records still need to be exported, tidied and imported, and that is still worth doing carefully.
And it does not make every business a fit. A large company with heavy compliance requirements and long-established processes may still need the depth of an established ERP. Unstack's plans also have a record limit for each app, so check that against your volumes before you commit.
The bottom line
For a business running on spreadsheets and disconnected tools, the old choice was to live with it or take on a long, expensive project. There is now a third option: describe what you need, look at what gets built, and adjust it as you go.
Start with the one process that hurts most. Describe it to the builder and see how far a single conversation gets you.