AI productivity tools have moved from personal experiments to company line items. About a quarter of the companies interviewed since 2025 that list their stack now name an AI assistant in it. The buyers rolling them out describe a decision made twice: once by IT and security, who pick a standard that fits the office suite and the data rules, and again by the teams using it, who often reach for a different tool because it writes, codes or analyzes better.
What is an AI productivity tool?
AI productivity tools are the AI assistants and agents companies give employees for everyday work: writing and drafting, research and summarizing, data analysis, coding and meeting notes.
The assistants buyers name are ChatGPT, Claude, Gemini and Microsoft Copilot, usually on business or enterprise plans, alongside tools such as Glean for searching company knowledge. The choice of standard is often settled by what the company already runs: “We use Copilot primarily because we are on the Microsoft suite.”
What the interviews add is how those tools are actually bought and used:
In these interviews the company standard follows the office suite and the data terms — Copilot at Microsoft companies, Gemini at Google ones — while teams choose on output quality, so about two in five companies with an assistant in their stack run more than one.
Rollout is governed closely and measured loosely: buyers describe AI councils, approved-tool lists and months-long approvals, but no company here reports a measured time saving. AI agents are the next step, and for most companies they are still a pilot or a plan.
What is the difference between an AI assistant, an enterprise AI plan and an AI agent?
An assistant answers a person's request; an enterprise plan is the same assistant under company data terms; an agent carries out a multi-step task with less supervision. Most companies here use the first two daily and are still working out the third.
The enterprise plan is what IT is really buying. Buyers who standardize describe security first: controlling which company data employees can share, how vendors process and retain it, and who can reach the tools through company-managed identities. One buyer states the line plainly: “I wouldn't be comfortable or want my teams to be putting information we have into a public commercial AI, like a ChatGPT, right, unless it was an owned instance.”
Agents are where the tools start doing work rather than helping with it, and where most of the plans and pilots sit.
AI assistant
A chat-based tool that drafts, summarizes, analyzes and codes on request — ChatGPT, Claude, Gemini or Microsoft Copilot.
Enterprise AI plan
An AI assistant licensed for organizational use, with administrative controls, identity management and contractual protections for company data. Training exclusions, retention policies and security features depend on the provider and plan.
AI agent
An AI that carries out a multi-step task or workflow, such as routing requests or querying company data, built in tools such as Copilot Studio, Glean or in-house.
Who decides which AI tools employees use?
IT, security or an AI council, more often than the team that will use the tool. About one in six companies discussing these tools names an AI governance committee, and approved-tool lists and written AI policies are common. Approval can be slow: “it's probably like a six-month process to onboard a new vendor or new capability, particularly if it involves AI,” says one buyer.
The standard tracks the company's software. Among companies interviewed since 2025, about three-quarters of those running Microsoft Copilot have more than a billion dollars in revenue, against fewer than half of all companies, and Copilot is about two and a half times as common at companies that already run Microsoft's business and data products. Claude, ChatGPT and Gemini spread more evenly across sizes, with Claude skewing toward smaller companies. A common result is one assistant for everyone and a second for a few: “We've given Microsoft Copilot to the entire organization. We have a small group using Claude.”
Where the sanctioned tool falls short, employees route around it — “Most people use apps on their phone like ChatGPT and then send it back to themselves, and it's kind of silly,” one buyer says — and several companies describe this kind of shadow AI.
What do employees use AI productivity tools for?
Five main activities, led by developing internal agents and automations — counting projects in development and pilot as well as agents already in daily use. Among companies describing their AI work, these activities appear in this order:
- Building internal agents and automations
- Writing and drafting
- Coding and software development
- Data analysis
- Research and searching company knowledge
Meeting notes and customer-facing content are smaller jobs.
Building internal agents and automations
Companies are building agents for knowledge lookups, ticket intake and data questions. About one in eight companies in this group reports agents in daily use, while others describe pilots or planned deployments.
One buyer describes a working example: “We've also built our own in-house MCP and agent combination to use natural language to query the data in our lakehouse too.”
Writing and drafting
Emails, documents, briefs and marketing copy.
As one puts it: “It also has allowed us to write copy and do a lot of things that it just took us days to do in the past.”
Coding and software development
Writing queries and code, and modernizing old code, often with a different tool from the company-wide assistant.
One buyer on a coding assistant: “We connected our GitHub to it and it helped us fix a lot of old bad code to modernize it.”
Data analysis
Working through spreadsheets, reports and data exports: “I also use our paid premium subscription to ChatGPT to help with a lot of the data analysis.”
Research and searching company knowledge
Summarizing documents and finding answers across internal systems. One buyer says a knowledge-search tool is “really helpful for being able to surface company knowledge and also be able to gather information from a lot of different sources across the business.”
Building agents is slow going for some. Builders include Microsoft Copilot agents, Glean agents, Claude and in-house tools, and buyers are candid about the pace — “we are building AI agents in-house, but that's taking a long time, honestly.”
Which AI productivity tools do buyers rate highest?
Claude rates 8.5, ChatGPT 7.9, Gemini 7.6 and Microsoft Copilot 7.5; Glean rates around 7. These are overall product ratings. Our reading, not a buyer finding, is that the gap partly reflects how each tool arrives: the company-wide standard is the tool most employees are given, while a second assistant is usually one a team chose for itself.
| Tool | Rating | What it means here |
|---|---|---|
| Claude | 8.5 | Named in the most company stacks; often licensed for a smaller group, praised mostly in comparison with other assistants, and criticized for token costs. |
| ChatGPT | 7.9 | Praised for speed and analysis; also the tool employees most often use outside sanctioned options. |
| Gemini | 7.6 | The standard at Google Workspace companies; praised for its data controls, criticized for long answers and sourcing. |
| Microsoft Copilot | 7.5 | The most common company-wide standard, chosen for the Microsoft 365 fit; most complaints compare it with other assistants. |
| Glean | around 7 | Too few buyers for a decimal. Search and agents across company knowledge. |
Assistants
Microsoft Copilot · ChatGPT · Claude · Gemini · Perplexity · NotebookLM
Knowledge, coding and agents
Glean · GitHub Copilot · Cursor · Copilot Studio · Zapier · Writer
What do AI productivity tools cost, and how do companies measure them?
Cost is counted in seats and tokens, and value is barely counted at all. No prices are published here.
Buyers describe paying per seat for assistants and per token for heavier use, and both add up: “it burns through tokens fast and it costs us a lot of money,” says one about a coding tool, while another calls the company assistant's high cost “a bit of a challenge.” Several companies limit licenses to the people who use them. Measurement lags behind. No company here reports a measured time saving; the metrics buyers name are seat utilization and token use, and several say adoption is shallow — “right now we're still struggling with people just using the basics, which is not where we need them to be.” Among the complaints, data security comes up most, ahead of cost, with accuracy and connecting the tools to company data next.
Want a review of your own stack?
Get my review →Should your company standardize on one AI productivity tool?
Choose an approved standard that fits your existing systems and data requirements, but do not assume one assistant will meet every team's needs. The companies in these interviews often deploy a default tool broadly while licensing additional assistants for specialist work such as coding, analysis or agent development.
Three things follow from the interviews; the recommendations are ours, not practices buyers describe. First, decide the data rules before the tool — what can go into an assistant, on which plan — since that is what governance committees spend their time on. Second, provide an approval path for specialist tools when a team can show a need the standard assistant does not meet, rather than leaving employees to work around company policy. Third, measure use beyond seat counts: pick a few tasks, compare before and after, and cut licenses nobody uses.
For AI in specific marketing jobs, see what buyers use AI content creation for; for AI acting on behalf of shoppers, what is agentic commerce; and for being found by AI assistants, what is answer engine optimization.
Common questions
What are AI productivity tools?
AI productivity tools are the AI assistants and agents companies give employees to help with everyday work: writing and drafting, research and summarizing, data analysis, coding and meeting notes. The assistants buyers name are ChatGPT, Claude, Gemini and Microsoft Copilot, usually on business or enterprise plans, alongside tools such as Glean for searching company knowledge.
What is the difference between an AI assistant and an AI agent?
An AI assistant responds to a person's request: draft this, summarize that, analyze this file. An AI agent carries out a multi-step task or workflow with less supervision, such as turning requests into tickets or querying company data in plain language. In these interviews most companies use assistants daily, while agents are mostly in pilots or plans.
Which AI assistant do companies use most?
It depends on how you count. In company tech stacks, Claude is named most often, followed by ChatGPT and Microsoft Copilot, then Gemini. As the official company-wide standard, Microsoft Copilot is the one buyers describe most, usually because it comes with Microsoft 365, and Gemini is the standard at companies on Google Workspace.
Why do companies use more than one AI assistant?
Because the company standard is chosen on the office suite and data terms, while individual teams choose on output quality. About two in five companies with an AI assistant in their stack run more than one. A common shape is one assistant licensed for everyone and a second for a smaller group, such as engineering or analytics, and some employees use unsanctioned tools on the side.
Who decides which AI tools employees can use?
Usually IT, security or a dedicated AI council rather than the team that will use the tool. In these interviews companies describe AI governance committees, approved-tool lists and written AI policies, with data security the most common concern. Approval can take months: one company describes about six months to onboard any new AI vendor.
How do companies measure whether AI productivity tools are worth it?
Mostly they do not yet. In these interviews no company reports a measured time saving from an assistant. The metrics buyers name are usage measures, such as seat utilization and token use, and several say adoption is shallow, with many employees using the tools only for basic tasks.
Are companies using AI agents yet?
Some are. About one in eight companies discussing these tools describes agents in daily use, for jobs such as internal knowledge lookups, ticket intake and querying data in plain language, and more are piloting or planning them. Builders include Microsoft Copilot agents, Glean agents and in-house agents, and some say building them is slow.
A tech stack review, powered by the buyers who've been there
What matters most to you, a shortlist for your stack, and the next steps to take — built from hundreds of verified buyer interviews.
Get my stack reviewFree · no password · your answers are anonymized before they inform research like this.