Developers and technical leaders spend 33% of their working time building and maintaining internal tools, according to Retool's 2022 State of Internal Tools survey of 2,285 people. That is the most quoted internal tools statistic, and it comes with a caveat that changes how you should read it: 87% of the respondents were active Retool customers, people who build internal tools for a living.
This page collects the numbers that exist on internal tools: engineering time, what the tools are built on, the size of the low-code market, the 2026 shift from buying SaaS to building, and the error rates of the spreadsheets that most internal tools replace. Each figure carries its source, its sample and, where the publisher sells the thing being measured, a label saying so.
How much engineering time goes to internal tools
Retool has asked the same question in four annual surveys. The answer has drifted upwards, and it climbs with company size until the very largest firms.
| Edition | Sample | Headline time figure | By company size |
|---|---|---|---|
| 2020 | 310 respondents, mostly 2 to 500 employees | More than 20% of team time at companies over 20 staff | Close to 40% at companies over 1,000 staff |
| 2021 | 650 developers and technical leaders, Feb to Mar 2021 | More than 30% of developer time | 45% at companies with 5,000 or more staff |
| 2022 | 2,285 respondents, May 2022, 87% Retool customers | 33% of respondents' time | Rises with size, dips slightly at 10,000 or more |
| 2023 | 2,276 respondents, March 2023 | About 86% say their company spent as much or more time than the year before | Increase reported at every company size |
Source: Retool State of Internal Tools 2020, 2021, 2022 and 2023
Retool sells an internal tools platform, so these are vendor-sourced figures from a vendor-weighted sample. A third of engineering time is plausible for a team that already uses a tool builder. It is probably high for a typical product engineering team, and we have not found an independent survey that measures the same thing.
Two other surveys come at the question from the side. MuleSoft's 2025 Connectivity Benchmark, a survey of over 1,050 IT leaders run by a company that sells integration software, found IT teams spend 39% of their time creating custom integrations such as automations. The average enterprise in that sample runs 897 applications and only 29% of them are integrated, which is the gap most internal tools exist to close.
Atlassian's 2025 State of Developer Experience survey of 3,500 developers and managers found 50% lose 10 or more hours a week to non-coding friction, and 90% lose six or more. Its write-up notes that developers spend only about 16% of their time actually coding. Internal tools are one of the places the other 84% goes, in both directions: building them takes time, and missing ones cost it.
Who the tools are for and what they sit on
The 2022 survey was the first in which operations overtook every other team as the main customer for internal tools.
| Data source | 2022 | 2023 |
|---|---|---|
| Internal databases | 82% | 78% |
| Internal APIs (REST, gRPC) | 63% | 56% |
| Third-party APIs (Stripe, GitHub, Slack) | 44% | 45% |
Source: Retool State of Internal Tools 2022 and 2023
Teams served. Operations 55%, customer support 52%, sales 39% and engineering 38% in 2022. A year earlier, customer support and engineering (41% each) narrowly led operations (40%).
Dedicated staff. 57% of 2022 respondents said their company has at least one full-time role dedicated to internal tools, most often in engineering (38%), then operations (20%) and data (17%). In 2021, 77% of companies with 500 or more employees had a dedicated internal tools team.
Stack. Among respondents building from scratch, 57% used React in 2022 and just over half in 2023. PostgreSQL was the most common database. 13% used an admin framework in 2022, rising to 18% in 2023, with Django Admin the clear favourite at 51% and then 50%.
The low-code market, by Gartner's numbers
Gartner's most recent public market forecast dates from December 2022. It put worldwide low-code development technology revenue at $26.9 billion for 2023, up 19.6%.
| Segment | 2021 | 2022 | 2023 | 2024 |
|---|---|---|---|---|
| Low-code application platforms (LCAP) | 6,324 | 7,968 | 9,960 | 12,351 |
| Integration platform as a service (iPaaS) | 4,680 | 5,668 | 6,668 | 7,838 |
| Robotic process automation (RPA) | 2,350 | 2,892 | 3,401 | 3,879 |
| Multiexperience development platforms | 2,081 | 2,508 | 2,999 | 3,563 |
| Business process automation (BPA) | 2,416 | 2,585 | 2,761 | 2,940 |
| Citizen automation and development platforms | 554 | 732 | 953 | 1,232 |
| Other low-code technologies | 92 | 109 | 126 | 146 |
| Total | 18,497 | 22,462 | 26,869 | 31,949 |
Source: Gartner, December 2022, as republished by TelecomTV
The same release predicted that by 2026, the year we are now in, developers outside formal IT departments would make up at least 80% of low-code tool users, up from 60% in 2021. InfoWorld's report of the forecast adds two more 2026 figures: low-code tools accounting for 75% of new application development, up from 40% in 2021, and a market of $44.5 billion. We could not find a public Gartner update that checks those forecasts against what happened, so treat them as forecasts, not results.
Gartner's 2024 Magic Quadrant for enterprise low-code application platforms, quoted by Mendix (a vendor in it), adds two newer predictions: 60% of software development organisations will use an enterprise LCAP as their main internal developer platform by 2028, up from 10% in 2024, and LCAPs will be used for mission-critical development in 80% of businesses by 2029, up from 15% in 2024.
The survey data is more modest than the forecasts. In Retool's 2021 survey only 38% of developers used low-code or no-code tools at all, though 88% of those planned to keep doing so.
Build, buy, or both
By 2023, the line between building and buying had blurred. Home-grown apps were still the majority (55% of Retool respondents), but nearly 65% also used something else: spreadsheets (48%), business process tools (35%) and BI or visualisation tools (27%). Among respondents using a third-party platform, about 65% used Retool, 36% Zapier, 32% Airtable and 26% Google AppSheet, figures that reflect who answered a Retool survey.
Retool's 2026 Build vs. Buy report, a survey of 817 Retool customers and builders conducted in late 2025, describes a further shift. 35% have already replaced the functionality of at least one SaaS tool with a custom build, and 78% plan to build more custom tools in 2026. 36% of respondents were software engineers; the rest worked in operations, product, data, IT, marketing, finance and other roles.
| Category | Share of respondents |
|---|---|
| Workflow automations | 35% |
| Internal admin tools | 33% |
| BI tools | 29% |
| CRMs and form builders | 25% |
| Project management | 23% |
| Customer support | 21% |
Source: Retool, 2026 Build vs. Buy report
AI is part of the reason. 51% of respondents have built production software with AI that their teams use, and about half of those (49%) say it saves six or more hours a week. Among those who have shipped, 72% use AI to write pieces of code they integrate themselves, and 31% prompt their way to complete apps. Only 8% use AI-generated code without changes, and 44% test it thoroughly before use.
Spreadsheets: the internal tool nobody scoped
Spreadsheets are where most internal processes start. Almost half of Retool's 2023 respondents use a spreadsheet to manage at least one internal process, with Google Sheets at 81% and Excel at 49% among them. In the 2020 survey, more than 40% said they used Google Sheets to build internal tools.
The research on spreadsheet accuracy is old but consistent.
| Finding | Sample | Source |
|---|---|---|
| Average cell error rate of 3.9% | 14 lab studies, 967 people working alone | Panko, EuSpRIG 2015 |
| Errors found in 94% of spreadsheets | 85 spreadsheets, intensive inspections | Panko, EuSpRIG 2015 |
| Errors found in 88% of spreadsheets | 113 spreadsheets audited 1995 to 2007 | Panko, as reported by Journal of Accountancy, 2015 |
| About one-fifth of papers with Excel gene lists contain gene name conversion errors | Leading genomics journals | Ziemann and others, Genome Biology, 2016 |
| 15,841 Covid cases left out of official figures | Public Health England, 25 September to 2 October 2020 | BBC, 2020 |
Source: Panko 2015; Journal of Accountancy 2015; Genome Biology 2016; BBC 2020
The per-cell rate looks small, but it compounds. Panko's point is that a 1% to 5% cell error rate across a chain of 100 dependent calculations makes a wrong bottom-line figure close to certain, and that people are poor at finding errors after the fact: in the software development studies he cites, error detection rates are only 20% to 40%.
The Public Health England case shows the failure mode internal tools are built to prevent. Testing labs sent clean CSV files. PHE's own process loaded them into Excel templates in the old XLS format, which holds about 65,000 rows, and the overflow was silently cut. Of 50,786 cases initially reported for that week, 15,841 were missing.
Shadow IT and access control
Governance has not kept up with building. In Retool's 2022 survey, over a quarter of respondents said their internal tools had no access controls. In the 2026 report, 60% had built something outside IT oversight in the past year and 25% did so frequently. Their reasons were speed (31%), unmet needs (25%) and IT's process being too slow (18%), and what they built was mostly internal tools (53%), automated workflows (53%) and custom dashboards (51%).
Gartner has put the wider figure at 41% of employees acquiring, modifying or creating technology outside IT's visibility in 2022, rising to 75% by 2027, as reported by the IEEE Computer Society.
The blockers respondents name are the same ones that slow any internal build. In Retool's 2026 report, the top technical blockers to automating more work were engineering bandwidth (42%), security and compliance (41%) and integration between systems (39%). The top organisational ones were unclear ROI (33%), budget (30%) and maintenance burden (26%).
How teams measure whether a tool worked
Employee productivity is the main yardstick: 54% in 2021, 60% in 2022 and 58% in 2023. In 2023, employee adoption (35%, up from about 25%) and internal user satisfaction (30%) followed, with reduced software costs at 30%, increased revenue at 25% and lower headcount needs at 21%. Some respondents said their company does not measure internal tools at all, and 35% of organisations in the 2026 report have no AI productivity metrics.
What this means for a team deciding whether to build an internal tool
Start from your own hours, not the 33% headline. Count the time a process takes each week, how many people touch it, and what an error costs, then compare that with the build and the upkeep. The surveys suggest three things worth acting on: the first candidates are operations and support workflows that run on a shared spreadsheet; access control and an owner should be in scope from day one, not after an incident; and a low-code builder is the fast route for standard admin screens, while custom code earns its place when the tool needs its own data model, complex permissions or a customer-facing surface later.
We build marketing sites in Webflow and applications in Next.js, so read our view with that bias in mind. When an internal tool outgrows a spreadsheet or a builder, we would scope it as a small web app with a clear owner. Our web apps service page explains how we quote that per project.
