Build Internal Tools with AI

Build vs buy internal tools in the age of AI builders

AI has made building cheaper, but not free. The evidence points to a middle option for most companies, and to a short list of cases where buying is still the better decision.

Three colleagues at a wooden meeting table compare two printed vendor proposals side by side, one pointing at a line on the page, a laptop beside them.
The build-or-buy decision is usually made on paper, one process at a time.
Contents
  1. Why the decision is harder now
  2. The three options
  3. When buying is the right decision
  4. When building on a platform is the right decision
  5. The platforms differ in what "build" means
  6. A worked comparison
  7. The path to a decision
  8. Questions buyers ask
  9. Sources

The old version of this decision was simple: buy software unless the process is so particular that nothing fits, and then ask engineering for a project. AI builders have changed the cost of the second option, which means the decision now has three branches instead of two.

Why the decision is harder now

Two pressures meet in the same budget meeting. On one side, rented software keeps growing: Zylo's 2026 index reports a median SaaS spend of $9,455 per employee and finds that organisations leave 36% of their licences unused⁠1. On the other, the promise that anyone can now describe a tool and have AI build it makes every renewal look negotiable.

The evidence on AI projects is a warning against building alone. In MIT's 2025 study, based on 150 interviews, a survey of 350 employees and 300 public deployments, purchased tools and vendor partnerships succeeded about 67% of the time, and internal builds about a third as often⁠2. The same study found the largest returns in back-office automation, which is exactly where most internal tools live.

The three options

OptionWhat you pay forWhat you ownWhere it fails
Buy a SaaS productSeats, usually per user per monthYour data, under the vendor's export termsThe process bends to the product; seats grow with headcount
Build from scratchEngineering time, hosting, maintenanceEverything, including every bugEngineers spend their time on tools instead of the product
Build on a platformA platform fee plus builder timeThe tool, and on some platforms the code and dataPlatform limits, plan tiers and lock-in

The cost of the second option is well documented. In Retool's 2022 survey of 2,285 developers and technical leaders, respondents reported spending 33% of their time on internal tools⁠3. That is a vendor's survey, but it explains why the third option has grown.

When buying is the right decision

  • The function is regulated or standard. Payroll, tax, the general ledger and statutory reporting encode rules that change every year. A mature product carries that burden for you.
  • The process is the same in every company. Email, video calls and document storage rarely differ enough to justify a custom tool.
  • Nobody will own the tool. A built tool without a named owner decays. If no one will maintain it, rent something that someone else maintains.

When building on a platform is the right decision

  • The process is yours. How you quote, schedule, approve or onboard is often the part of the business that differs from competitors.
  • The data already lives in a spreadsheet or a database you control. The tool adds structure and permissions rather than migrating to a vendor's model.
  • Seat costs grow faster than value. When most users only read or update a few records, per-seat pricing is the wrong shape. Airtable, for example, bills only for users with edit rights, and read-only collaborators are free⁠4.

The platforms differ in what "build" means

The platforms in our main comparison sit at different points between buying and building. Appsmith and Retool give developers components and data connections, with AI assisting the build⁠56. Softr and Airtable let non-developers assemble tools on their own databases⁠78. Lovable writes a full web app from a chat and hands over the code⁠9.

OpenLazarus sits closest to buying a whole system while still owning it. Its builder designs a CRM, mail, operations and finance system around the company and lists the subscriptions the design would replace⁠10. The code and the data belong to the company, and hosting can move to its own cloud⁠1112.

Three questions that settle most cases

  • Is the process regulated or identical everywhere? Buy.
  • Will a named person own the tool for the next two years? If not, buy.
  • Do most users only update a handful of records? Build on a platform, and compare per-seat and per-account pricing.

A worked comparison

Take one function that many companies rent: a CRM used by 25 people. The comparison is not the subscription price alone. Add the hours spent copying data between the CRM and the spreadsheets around it, the integrations that keep them in sync and the seats that go unused. Then price the platform alternative for the same 25 people, using the list prices in our main guide, and add the hours of the person who will maintain it. If the platform total is lower and the person exists, building is the stronger case.

The path to a decision

  1. List the five tools or spreadsheets that cost the most time. Use invoices for software and a week of time notes for spreadsheets.
  2. Sort each into buy, build or build on a platform using the three questions above.
  3. Name an owner for every tool you plan to build. No owner, no build.
  4. Price two platforms for the first build at the plan that includes the access controls you need, not the entry plan.
  5. Build one tool, retire one subscription, and measure for a quarter before starting the second.

The end state is a short, deliberate stack: bought products for regulated and standard work, and owned tools for the processes that make the company different, each with a named owner and a known monthly cost.

For the security side of this decision, read what to set up before an internal tool goes live.

Questions buyers ask

Is it cheaper to build or buy internal tools?

It depends on how many people use the tool and how particular the process is. Per-seat software becomes expensive as headcount grows and when licences go unused; Zylo's 2026 index puts unused licences at 36% on average. Building on a platform costs a monthly fee plus the time of whoever maintains the tool.

Do AI app builders make building from scratch a good idea?

Rarely for a company without engineers. AI writes the first version quickly, but someone still has to own access control, data connections, updates and support. Building on a platform that provides those pieces carries less risk than maintaining generated code alone.

When should a company always buy?

For regulated or standardised functions where a mature product already encodes the rules, such as payroll, tax filing and the general ledger. Internal tools work best around those systems, not in place of them.

Sources

  1. 2026 SaaS Management Index. Zylo, 29 Jan 2026. Accessed 8 Nov 2026.
  2. MIT report: 95% of generative AI pilots at companies are failing. Fortune, 18 Aug 2025. Accessed 8 Nov 2026.
  3. State of Internal Tools 2022. Retool, 1 Feb 2023. Accessed 8 Nov 2026.
  4. Pricing. Airtable. Accessed 8 Nov 2026.
  5. Appsmith: build custom internal apps. Appsmith. Accessed 8 Nov 2026.
  6. Build, deploy and manage internal tools. Retool. Accessed 8 Nov 2026.
  7. Pricing. Softr. Accessed 8 Nov 2026.
  8. Airtable AI. Airtable. Accessed 8 Nov 2026.
  9. Pricing. Lovable. Accessed 8 Nov 2026.
  10. Design your company’s operating system. OpenLazarus. Accessed 8 Nov 2026.
  11. Frequently asked questions. OpenLazarus. Accessed 8 Nov 2026.
  12. Pricing. OpenLazarus. Accessed 8 Nov 2026.

Updated 8 Nov 2026: First published.