What AI Should (and Shouldn't) Do in Client Work
AI should take on the repetitive work that surrounds a client relationship: drafting, summarizing, organizing, and watching for risk. People should keep the work that defines the relationship: judgment, commitments, and trust. The firms that get this split right will serve more clients with the same team, without their clients ever feeling handled by a machine.
The Simple Rule
A useful test for any client task: if getting it wrong would change what the client believes about you, a person owns it. AI can prepare the work. A person decides what reaches the client.
What AI Should Handle
Organized by the four pillars of client management:
Intake
Turning a messy brief or call transcript into a structured first draft of scope
Flagging missing information before kickoff, like approvers, deadlines, or budget
Pulling context from the sales process into the client record
Delivery
Converting client feedback into specific, assigned tasks
Drafting first versions of routine deliverables for human review
Spotting capacity conflicts across accounts before deadlines slip
Communication
Summarizing client meetings and capturing action items
Drafting weekly status updates from actual project data
Answering "where is this?" questions from a shared source of truth
Account Health
Watching for early risk signals: slower responses, rising revisions, budget burn
Preparing account summaries before renewal conversations
Surfacing patterns across the portfolio that no single account lead would notice
What Should Stay Human
Scope and pricing decisions. AI can draft the change request. A person decides whether it's billable.
Bad news. Delays, mistakes, and budget overruns are delivered by a person, ideally live.
Relationship repair. When a client is unhappy, the response has to come from someone accountable.
Strategic advice. Clients pay for your judgment. AI can inform it, not replace it.
Final approval of anything client-facing. At least until your firm has clear evidence a specific output type is reliable.
Set Review Rules Before You Scale
The risk isn't that AI drafts something wrong. It's that a wrong draft reaches a client because nobody owned the review. Before rolling AI out across accounts, write down:
Which outputs can reach clients, and which are internal only.
Who reviews each type, by role, not "whoever is around."
What never gets automated, no matter how good the draft looks.
How clients are told, if at all, that AI assists your team.
When the rules get revisited, such as quarterly, as your confidence in specific outputs grows.
AI Amplifies the System You Already Have
AI works from your client records, project data, and meeting notes. If projects aren't tied to the right clients, if scope lives in email, or if status is tracked in someone's head, AI will produce confident output built on bad information. Fix the underlying client management system first, and AI makes it faster. Skip that step, and AI makes the mess move faster too.
The Bottom Line
Use AI to remove the busywork around client relationships, not the relationships themselves. Automate the drafting, summarizing, and monitoring. Keep people accountable for judgment, commitments, and anything a client will remember.
ClientManagement.ai is an independent resource on client management practice and software.