What should a customer success leader do about the AI tools their team already uses?
AI is probably already part of your customer success team's work, whether or not anyone bought it. A customer success manager finishes a call, pastes the transcript into a general-purpose AI chat tool and asks for a follow-up email. Before a renewal, an account manager copies usage figures from one dashboard and support tickets from another into the same chat window and asks for a summary. The work gets done faster, none of it shows up in the CRM, and the leader usually has no idea it happened.
TSIA's State of Customer Success 2026 report describes the data problem for AI in customer success as one of access: the AI a team uses "only sees fragments" of the customer. A personal chat window is the clearest example. It sees what one person chose to paste, for one account, on one afternoon.
The best response is to treat that private use as a map. Find out, without blame, which account work people already hand to AI, because that list shows which work is worth handing over properly. Then move that work out of personal chat tools and into defined work that runs on each account's full data, writes its results back to the account record and stays inside the company's controls. Banning the tools loses the map, and ignoring them leaves the gains small and invisible.
How much AI your team already uses without telling you
Nobody has measured this for customer success. The research behind this piece found no independent study of how customer success managers, account managers or renewals managers spend their time, let alone how much of that time goes into AI tools. Studies across the whole workforce all point the same way.
KPMG and the University of Melbourne surveyed more than 48,000 people for their 2025 study of trust in AI and found that 57% of employees hide their AI use. Separately, MIT's NANDA research found workers using personal AI tools at more than 90% of companies, while only 40% of companies had bought an official subscription.
Customer success work suits this kind of private use, because so much of it is reading and writing about accounts: call notes, follow-ups, meeting preparation and renewal summaries. If your team looks like the wider workforce, some of its members are likely to be using AI on that work already, and many of those would keep quiet about it if asked in a team meeting.
What the AI tools your team already uses tell you
The tasks people choose to give to AI on their own tend to be tasks they find repetitive, where they trust the output enough to use it. That makes them a better starting list than any vendor demo, because the list comes from the people doing the work.
Getting the list means asking in a way that makes honesty safe. A leader who asks the team whether anyone is pasting customer data into chat tools will probably hear silence. A short anonymous survey works better, with four questions:
- Which account tasks do you already use AI for? Ask for the task in plain words, such as "the follow-up after a customer call", so the answers can be grouped.
- Which tool do you use? This tells you where customer information is going today, which is something the company should know.
- What do you paste in? The answer shows which systems each task draws on, such as call transcripts, CRM notes, usage exports or support tickets.
- How often do you fix what comes back? Work that people already check and correct by hand is work whose output can be checked, which is one of the things that makes a task safe to hand to an agent.
In customer success, the likely candidates are tasks like post-call follow-ups, call notes, meeting preparation and renewal summaries. Each one is frequent, each one depends on information that already sits in company systems, and each one produces something a person can check before it reaches a customer.
Why personal AI chat tools hit a ceiling on account work
Private use deserves some credit. Wharton and GBK's 2025 survey of about 800 decision-makers found that roughly three in four report a positive return on AI, mostly from productivity uses like these. The limits appear when the same tools are asked to do account work, and three of them matter most.
- A chat tool only sees what someone pastes in. Each conversation starts with whatever the person copied across: one transcript, one export, one email thread. The CRM history, product usage, support tickets and billing data that explain the account stay where they are unless someone fetches them. TSIA also found that the ability to unify customer data is one of the strongest predictors of renewal performance, which is exactly what a single pasted fragment lacks.
- Pasting in the whole history makes the answers worse. The obvious workaround is to paste in everything about the account. Research on long inputs by Chroma, a company that sells AI data infrastructure, found that all 18 leading models it tested gave worse answers as the input grew, and that irrelevant material hurt more than length alone. An agent working on an account needs the parts of its history that matter for the task, selected each time, and a person pasting under time pressure rarely has the time to select them.
- The work leaves no trace in the account record. A follow-up drafted in a personal tool goes out by email, but the next steps it contains reach the CRM only if someone copies them back by hand. The colleague who inherits the account cannot see what was prepared or promised, and the company cannot see which customer data went into which tool.
Is AI chat use delivering measurable productivity results?
The largest study of this kind of use so far suggests the gains stay small. A National Bureau of Economic Research working paper by Humlum and Vestergaard (2025) followed about 25,000 workers in Denmark who use AI chatbots in jobs that were otherwise organized the same way. Average time savings were only 3%, and the authors found "precise null effects on earnings and recorded hours."
That study covers the Danish workforce as a whole and says nothing specific about customer success. Its lesson still applies to a customer success team. A few minutes saved on each follow-up email, inside a week that is otherwise unchanged, come back in small pieces spread across the team. Nobody directs that time anywhere, so the accounts that were getting too little attention before keep getting too little attention.
How Trig turns private AI use into account work the whole team can see
Trig is an agent-powered workspace for account management and customer success. AI agents do recurring account work, and the team directs and reviews it. Work in Trig is defined as an assignment, which sets out what to produce, which accounts to work on, who receives the work, when it runs, and what agents may read and do. Before an assignment runs, the team reviews Trig's plan: the information it will use, the actions it will take and the work it will produce.
Agents work from a profile of each account that combines CRM, product, support and billing data with conversations and previous agent work, and that profile updates as new information arrives. Completed work goes where the team already works. Briefs and reports can be read in Trig or delivered to a connected messaging or document tool, and notes, next steps and field updates are written to the CRM record. Every run is recorded, so anyone can trace a brief or an update back to the assignment that produced it. Customer data is never sent to external AI providers, and prompts and responses are never used to train foundation models.
The tasks a survey like the one above tends to surface match templates Trig already includes, such as drafting the follow-up with next steps after each customer call, updating the CRM with call notes, preparing a brief for every customer meeting and assembling a renewal report two weeks before each renewal.
What this looks like in practice
Consider a customer success team of six people covering 450 accounts. The team lead runs the four-question survey anonymously. Four people say they draft post-call follow-ups in a personal chat tool most days by pasting in their notes. Three build renewal summaries the same way, copying usage from the product dashboard and tickets from the support tool. Two mention that they often have to fix next steps the tool made up.
The team lead turns those two tasks into two assignments. The first drafts a follow-up email with the agreed next steps after each customer call, leaves it in the account owner's drafts, and writes the next steps, risks and new contacts to the account record. The second assembles a renewal report two weeks before each renewal, with contract details, usage, support history and call summaries, for every renewing account, including the ones nobody had time to prepare.
The team lead reviews both plans before they run and checks the first few reports line by line. The follow-ups stay as drafts for each owner to edit and send. People keep using chat tools for whatever else they find them useful for, and the two tasks that touch customer data every day now run on the full account record, in a place the team lead can see. These figures are an example, and every team's survey will turn up its own list.
What this means for the customer success team
For a customer success leader, the AI your team already uses is the most honest signal available about which account work is ready to hand over. Asking about it openly turns a private habit into a shared list. Moving that list into defined work that runs on each account's full record turns minutes saved by individuals into work done for every account, recorded where the next person to look after that customer can find it.







