.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
REPORT
2025 healthcare email security report
Key insights from 180 email-related healthcare breaches and actionable steps to protect your organization.
REPORT
2025 healthcare email security report
Key insights from 180 email-related healthcare breaches and actionable steps to protect your organization.
What the benchmark found
Healthcare organizations adopted AI coding assistants to move faster.
The builder pool widened at the same time. IT staff build with these tools at 84% of organizations, developers at 58%, and at two in five, developers are not involved at all.
Every figure describes healthcare organizations that already use AI coding assistants, not healthcare organizations generally. Survey of 151 healthcare leaders, fielded July 2026.
- Nine in ten describe AI adoption as formal and organization-level.
- 80% have put patient-facing AI-assisted tools into production, ahead of internal scripts and automations at 46%.
- 82% connect systems to AI through APIs; 79% report comparable use of MCP servers.
Downloads
The full report, plus standalone cuts sized for different audiences.
The 2026 benchmark for building with AI in healthcare
Who is building software with AI coding assistants inside healthcare organizations, what they build with them, the API and MCP infrastructure it runs on, and what the trajectory implies for review and governance capacity over the next year.
Get the reportExecutive summary Healthcare changed who counts as a developer
The full argument in a single page: who is building, what reaches patients, and what it runs on. Includes the By the numbers panel with all six headline figures.
Download the executive summaryInfographic Five numbers from the benchmark
Volume, adoption, patient-facing output, integration, and trajectory, each sized for sharing on its own spread. Built for slides and social posts where the full report will not fit.
Download the infographicExcerpt Adoption is higher than most estimates
Nine in ten describe adoption as formal and organization-level, and roughly three quarters built six or more tools in the past year. Only 4% of health systems have scaled AI with measurable outcomes.
Download excerptExcerpt IT staff are building more than developers
IT staff build with these tools at 84% of organizations against 58% for developers, and at two in five, developers are not involved at all.
Download excerptExcerpt AI-built software has already reached the patient
Patient-facing forms and portals sit at 55% and chatbots at 53%, both ahead of internal scripts at 46%.
Download excerptExcerpt AI is wired into the systems that run the business
Among these organizations, 82% connect systems to AI services through APIs and 79% report comparable use of MCP servers.
Download excerptExcerpt What comes next
Seventy percent expect AI-assisted development to increase over the next 12 months, and 1% expect a decrease. Three things worth planning around, plus what Forrester and Gartner project outside healthcare.
Download excerptCharts
The four figures behind the findings, each with its own base and scale. Percentages come from the survey of 151 healthcare leaders, fielded July 2026.
Who uses AI coding assistants to build
Share of organizations reporting each group. Multiple selections allowed. n=151
Source: Paubox, The 2026 benchmark for building with AI in healthcare.
What these organizations have built with AI assistance
Share of organizations reporting each output. Multiple selections allowed. n=151
Source: Paubox, The 2026 benchmark for building with AI in healthcare.
Why these organizations use AI coding assistants
Share rating each reason very or somewhat important. n=151
Source: Paubox, The 2026 benchmark for building with AI in healthcare.
Number of tools built with AI assistance in the past 12 months
Share of organizations. n=151
Source: Paubox, The 2026 benchmark for building with AI in healthcare.
Statistics
Every figure in the report. All describe healthcare organizations that already use AI coding assistants, not healthcare organizations generally. n=151.
Who builds with AI coding assistants, 2026
| Group | Share of organizations |
|---|---|
| IT staff | 84% |
| Developers or engineers | 58% |
| Operations or administrative staff | 53% |
| Compliance or privacy staff | 45% |
| Clinical staff | 38% |
| Executive leadership | 36% |
Why these organizations use AI coding assistants, 2026
| Reason (very or somewhat important) | Share of organizations |
|---|---|
| Increase development speed | 91% |
| Keep pace with innovation | 89% |
| Address limited engineering resources | 85% |
| Enable non-developers to build solutions | 81% |
| Reduce costs or staffing needs | 80% |
What these organizations build with AI assistance, 2026
| Output | Share of organizations |
|---|---|
| Data integrations or APIs | 65% |
| Reporting or analytics tools | 56% |
| Patient-facing forms or portals | 55% |
| Chatbots or virtual assistants | 53% |
| EHR integrations | 48% |
| Internal scripts or automations | 46% |
| Other third-party integrations | 13% |
Tools built with AI assistance in the past 12 months, 2026
| Tools built | Share of organizations |
|---|---|
| None | 1% |
| 1 to 5 | 17% |
| 6 to 20 | 39% |
| 21 to 50 | 30% |
| More than 50 | 7% |
| Don't know | 6% |
Adoption, infrastructure and outlook, 2026
| Measure | Share of organizations |
|---|---|
| Describe adoption as formal and organization-level | 88% |
| Connect systems to AI through APIs, at least moderately | 82% |
| Report comparable use of MCP servers | 79% |
| Have deployed patient-facing AI-assisted tools into production | 80% |
| Built six or more tools in the past 12 months | 76% |
| Rate enabling non-developers an important reason for adoption | 81% |
| Expect AI-assisted development to increase in the next 12 months | 70% |
| Expect AI-assisted development to decrease | 1% |
Sample composition, 2026
| Respondents | Share of organizations |
|---|---|
| Hospitals and health systems | 63% |
| Organizations with 201 to 5,000 employees | 76% |
| Total respondents | 151 |
| Margin of error at the 95% confidence level | +/- 8.0% |