Lawyers think they care; clients don't feel it. The gap is communication, not sincerity. Here's how AI buys back time without making messages feel robotic.
Seventy-two per cent of attorneys say "caring" is essential to client relationships. Only 40% of clients perceive that care equally. That gap - between how much you think you care and how much your clients actually feel it - is where most of the damage happens. And it has almost nothing to do with whether you genuinely care. It has everything to do with how, when, and how often you communicate.
I bring this up because the conversation about AI in client communication has become oddly binary. On one side, you've got the enthusiasts who think you can automate your way to better relationships. On the other, you've got the sceptics - and if you're an operations leader or a managing partner at a mid-market B2B service firm, you're probably closer to this camp - who hear "AI-assisted client communication" and immediately picture a robot sending stilted emails to your most important clients.
Our client relationships are built on personal contact. If clients feel they're communicating with a machine, we lose the trust that makes those relationships worth having.
That's a completely reasonable concern. And honestly, if the choice were between a personally crafted message from someone who knows the client and an AI-generated one fired off without oversight, I'd pick the first one every time. So would you. So would your clients.
But that's not the choice. The real choice, the one most firms are actually facing, is between "a fee earner who means to send that update but doesn't get round to it until Thursday" and "an AI-drafted update that the fee earner reviews, adjusts, and sends on Monday." Between a thoughtful response that arrives three days late and a thoughtful response that arrives the same afternoon.
When you frame it that way, the personal touch argument starts to work in AI's favour, not against it.
A few months back I was working with an operations director at a 200-person consulting firm. She'd been tracking how long it took her team to send routine project status updates to clients. The answer was, on average, about 45 minutes per update. Not because the updates were complex - most of them covered the same ground every fortnight - but because each one required pulling information from three different systems, drafting something coherent, and then second-guessing the tone for ten minutes before hitting send.
I remember her saying, "The irony is that the clients who most need a timely update are the ones whose updates take longest, because there's more to pull together." She wasn't wrong.
They started using AI to generate first drafts of those updates. The relationship manager still reviewed every one, tweaked the tone, added the personal observations that only someone who'd been in the room could add. Drafting time dropped from 45 minutes to about 12. That's not a small efficiency gain. That's the difference between updates going out on time and updates going out late - or not at all.
This is where AI genuinely earns its place. There are a handful of areas where I've seen it work well, and they all share the same logic: AI handles the structural and administrative dimensions of communication so that humans can spend more time on the relational dimension.
Drafting routine updates is the obvious one. AI-generated first drafts of project status updates, onboarding communications, standard advisory summaries - the fee earner personalises and approves. Across the engagements we've run at Distinction, this typically reduces drafting time by 60-80%, while keeping the human review that ensures accuracy and appropriate tone. The key word is "first draft." Not "final output."
Personalising content at scale is trickier but genuinely valuable. If you've got 150 clients and you want to send a relevant market update, AI can use client-specific data to adjust emphasis, surface the information most relevant to each client's situation, and tailor the communication - without requiring someone to write 150 individual emails. The communication still goes through human review. But the starting point is already 80% of the way there.
Summarising complex information is the underrated one. AI-generated plain-language summaries of complex legal, financial, or analytical documents help clients understand what they're receiving before they engage with the detail. I worked with a mid-market accountancy firm - around 80 staff, mix of owner-managed and PE-backed clients - who started adding a two-paragraph AI summary to the top of their tax advisory reports. Reviewed by the partner before it went out, always. Client feedback shifted noticeably within a couple of months. People felt more confident engaging with the full report because they understood the headlines first. One client told the partner it was "the first time I've actually read the whole thing."
Scheduling and logistics - meeting coordination, reminder sequences, follow-up nudges - are genuinely routine, genuinely welcome, and genuinely a waste of a senior person's time to manage manually. Let AI handle them.
And translation and accessibility are worth mentioning if you work across borders or if your client contacts include people outside your immediate area of expertise. A CFO receiving a legal update, for instance. AI translation and readability editing for non-specialists can make a real difference to whether the communication actually lands.
Right. So that's the good news. Here's the bit that matters just as much.
I had a conversation with the managing director of a financial services firm who'd rolled out AI-assisted communication across his client services team. On paper, it was working brilliantly - response times were down, consistency was up, the metrics looked great. Then he lost two clients in the same month. Both gave the same feedback: "It doesn't feel like you know us anymore."
What had happened was predictable, in hindsight. The team had started letting AI handle increasingly sensitive communications without enough human oversight. A client going through a restructure - a messy one, with redundancies and a board change - received a status update that was technically accurate and completely oblivious to what was actually happening in that business. The update referenced "the next phase of delivery" while the client was in the middle of deciding whether to continue the engagement at all. Nobody had read it before it went out. The client's response was two sentences. The MD read them to me and said, "That's when I knew we'd broken something."
So let me be specific about where AI communication assistance creates genuine risk.
Generic automated messages sent to clients who have a significant ongoing relationship with your firm. The client who receives a clearly automated acknowledgement of a sensitive query has learned something about how you prioritise their relationship. They won't forget it quickly.
Impersonal responses to emotionally charged situations. A client whose matter has gone wrong, whose financial situation has shifted, or who has raised a complaint needs a human response. AI-drafted responses in these contexts range from inadequate to actively damaging. A poorly timed automated message can turn a recoverable client relationship into a lost one - I've seen it happen, and it's not a slow process. It's a door closing.
And then there's the subtle one: communication that is correctly formatted, accurately informative, and obviously not written by someone who knows the client. It's technically fine. It just feels off. Your clients can tell. They might not articulate it as "I think AI wrote this," but they'll feel the absence of the small details - the reference to last month's conversation, the slightly informal sign-off, the thing that signals "I actually thought about you before I wrote this."
Any communication sent to a client without a named human reviewing and taking responsibility for it is, in my view, non-negotiable. Full stop.
I want to be really clear about this, because it's the spine of everything I'm arguing here.
In a professional services context where client relationships are your primary commercial asset, no AI-generated communication should reach a client without a human review. Not because AI drafts are always inadequate - they often aren't. They're frequently pretty good. But because the human review is what maintains accountability. It catches contextual errors that AI can't see. And it keeps the fee earner or relationship manager connected to the relationship in a way that matters.
The moment you remove the human from the loop, you've made a philosophical shift - from "AI helps me communicate better" to "AI communicates on my behalf." Those sound similar. They are worlds apart.
But we'd never actually remove the human review. That's just good practice.
Here's the thing: most firms say this, and most firms gradually stop doing it. Not through any deliberate decision - through drift. The workflow gets busy, the AI drafts are consistently decent, someone skips the review once and nothing bad happens, and within six months the "human in the loop" is a principle that exists in the onboarding documentation and nowhere else. The MD I mentioned earlier didn't decide to remove oversight. It just... eroded.
So the practical implementation has to be designed to resist that drift. AI drafts. A named human reviews, adjusts where needed, and approves before sending. The time saving comes from drafting time reduction, not from removing human involvement from the final output. If your workflow doesn't make that review genuinely easy - thirty seconds, not five minutes - it won't survive contact with a busy fee earner.
This is where most firms stumble. Not on the principle - everyone nods along with "human in the loop" - but on the implementation. Because if the workflow is clunky, fee earners won't use it. And if fee earners don't use it, you've bought a tool that sits there gathering digital dust.
Tool selection matters more than most people expect. AI writing tools that work within your existing communication platforms - Microsoft Copilot within Outlook and Teams, AI features within your CRM, or standalone tools that export to email cleanly - get adopted. Tools that require people to open a separate application, paste text back and forth, and fiddle with formatting don't. I know that sounds obvious, but you'd be surprised how many firms buy a brilliant AI writing tool and then wonder why nobody uses it three months later. The answer is almost always that it lives outside the workflow.
The AI drafting step should replace the blank-page moment, not add an extra step before it. Fee earners and relationship managers should be drafting with AI as naturally as they currently type. If it feels like an additional process, adoption will be low. If it feels like a faster version of what they already do, it'll stick.
The quality check before sending should take thirty seconds, not five minutes. Three questions: Is the tone right? Does this reflect what I actually want to say? Is there anything in the client's recent history that should change this? Frame it internally as the professional standard you'd want to apply regardless of whether AI was involved - not as "AI oversight" - and you'll get less resistance.
And then there's the one almost everyone skips: talking to clients. Not a survey. A conversation. "How are you finding our communication at the moment? Anything you'd change?" Two or three key clients, once a quarter. The most reliable signal that your workflow is working as intended comes from the people receiving the output, not the people producing it. If you're not doing this, you're flying blind - and the first signal you'll get that something's wrong is a client who's already decided to leave.
If you're going to introduce AI-assisted communication, you need to know whether it's making things better or just making things faster. Those aren't the same.
Response time improvement is the most straightforward metric - the reduction in time from trigger event to client communication. Measure it before and after. In our experience, a well-designed workflow typically moves average response times from two to three days down to same-day or next-day for routine communications. If you're not seeing something in that range, something's wrong with your workflow design.
Client satisfaction with communication quality is the one that actually matters. Measured through structured periodic feedback or through NPS questions specifically about communication. This is what tells you whether faster also means better. I'd suggest running a baseline before you introduce AI assistance - otherwise you're measuring against memory, which is unreliable.
Communication consistency across the client base is worth tracking too. Are all clients receiving communication that meets your firm's quality standard, or just the ones whose relationship manager happens to have more time? AI-assisted drafting should level this up. If it doesn't, you've got an adoption problem.
Adoption rate among fee earners is the earliest warning signal. What proportion of relevant communications are going through the AI-assisted workflow? Low adoption - say, fewer than half your fee earners using it after three months - almost always means the workflow creates more friction than it removes. The problem is the design, not the people.
The firms that get AI communication right aren't using it to communicate more cheaply. They're using it to communicate more often, more consistently, and more thoughtfully - because AI has taken the administrative burden off the drafting process and given that time back to the human who actually knows the client.
The firms that get it wrong are doing the opposite. They're using AI to remove humans from the communication process, and producing messages that are faster, more consistent, and less personal than the ones they replaced. That is genuinely the worst possible combination in a relationship-dependent business.
The enemy of personal communication was never technology. It was always time - and the slow erosion of good intentions under the weight of a full inbox. AI doesn't fix that automatically. But designed properly, it gives you a fighting chance.
If you want to understand which of your current client communication workflows are the best candidates for AI assistance - and how to design the human-in-the-loop process that maintains quality - book a client communication workflow review. We've also put together a client communication AI readiness checklist that covers the implementation criteria I've outlined above - it's a useful starting point before you commit to any tooling.