If you run an agency on your own, the maths is brutal. You're the strategist, the account manager, the producer, the bookkeeper and the person who remembers to send the invoice. Every hour spent writing a status email is an hour you're not selling, delivering, or resting. And what kills solo agencies usually isn't the work itself. It's the admin tax wrapped around the work.
So this post is about where Claude genuinely earns its keep in a one-person agency, and where it doesn't. I run several projects in parallel and I've tested most of this on my own projects, so I'll be specific about both sides. The aim is to get a couple of hours a week back without producing the kind of generic output that makes a client quietly start shopping around.
The split that matters: drafting versus deciding
Before any tooling, get one distinction straight. It determines everything else.
AI is brilliant at the mechanical layer of agency work. Aggregating data, restructuring it, drafting the first version of a recurring document, formatting, summarising, reformatting the same thing for a different audience. It's fast, tireless, and never gets bored on the eleventh client report of the month.
It is not your judgment. The diagnosis ("traffic's up but the wrong traffic"), the framing ("here's why that's actually fine this quarter"), the recommendation ("kill this campaign, double the other") — that's the bit clients pay a human for. Hand that over and you've turned yourself into a commodity. A client getting commodity advice will eventually wonder why they aren't just prompting the model themselves.
Keep that line visible. Everything below sits firmly on the mechanical side.
Client onboarding: the best place to start
Onboarding is the single best automation target for a solo agency. It's repetitive, it's front-loaded with admin, and a clunky version of it can cost you the relationship before you've delivered anything.
The mechanical parts Claude handles well:
- Turning a messy discovery call transcript into a structured kickoff brief — goals, scope, deliverables, who owns what, and the questions you still need answered.
- Drafting the welcome sequence: the "here's what happens next" email, the access-request checklist (analytics, ad accounts, CMS logins), the intake questionnaire.
- Producing a first-draft project plan with milestones from your standard process, so you're editing rather than staring at a blank page.
Picture the before and after. Onboarding that used to eat a chunk of your first week — writing the brief from scratch, assembling the welcome pack, chasing yourself to remember every login you need — becomes a 30-minute review of drafts. You feed Claude the call notes and your standard checklist. It returns the brief, the email, and the access list. You correct what's wrong and send.
The part you keep: deciding what this specific client actually needs, and reading the relationship. The brief is drafted by AI. Whether this engagement is scoped sensibly is your call.
Status reports: kill the blank page, not the thinking
Status updates are where solo agency owners quietly lose evenings. A client expects a regular update. You've done the work. Writing it up feels like unpaid overtime.
The pattern that works: give Claude the raw inputs — what got done this week, the numbers (pull them yourself or via a connected tool), anything blocked — plus a template that matches how you communicate. It drafts the update in your structure and a consistent tone. You add the one or two sentences of interpretation the client is actually paying for: what the numbers mean and what you're doing about it.
One caveat, and it's an important one. Don't let the model write the interpretation. A status report that says "engagement increased 12% week-on-week" with no human read on whether that's good, expected, or a fluke is exactly the kind of hollow update that erodes trust. The model assembles the facts. You supply the "so what".
If you want this to feel branded rather than ad-hoc, a consistent report format — same sections, same voice, every time — does a surprising amount of work for how senior you look. That consistency is genuinely easy to automate. The substance still isn't.
Capacity planning: the spreadsheet you keep avoiding
This is the least glamorous use, and possibly the most valuable. Solo agency owners chronically over-commit because they don't have a clear picture of their own bandwidth, then deliver late or burn weekends. Solo operators report more stress than people with teams for a reason. There's no one to absorb the overflow, so the overflow becomes your evenings.
Where AI helps: lay out your committed hours per client, your delivery deadlines, and your realistic working week, then have Claude model whether a new project actually fits before you say yes. Ask "if I take this on at 10 hours a week, what breaks?" and get a clear answer instead of optimistic vibes.
This is also where you catch the trap most coaches and consultants fall into. Running flat out at roughly 80% capacity with no slack means any wobble — a sick day, a client emergency — tips you into overrun. Seeing that on paper is what lets you say no, or raise your rates, before you're underwater. The model does the arithmetic. The decision to protect your capacity is a business choice only you can make.
Retainer reviews: the conversation you keep postponing
Retainers drift. The scope creeps, the original deliverables stop matching what you actually do, and because the money arrives every month you never sit down to reconcile it. That's revenue and goodwill leaking quietly.
A quarterly retainer review is the fix, and AI makes the prep painless. Feed it the original agreement, what you've actually delivered over the period, and the hours it took. It'll surface the gaps. Where you're over-delivering (a rate conversation). Where a deliverable has gone stale (drop it or swap it). Where the relationship has outgrown the scope (an upsell, honestly framed). It can draft the review document and the talking points for the call.
The part that stays human is the call itself, and the read on whether this client wants more, wants less, or is a flight risk you should handle gently. AI preps the review. You have the conversation.
White-label deliverables: useful, with a contract caveat
If you produce work that goes out under a client's brand, or you subcontract, AI speeds up the formatting and first drafts of recurring deliverables — monthly performance summaries, audit documents, the recurring report that always looks the same but takes an hour to assemble. Strip your branding, apply theirs, done.
Two real cautions here, because this is where solo operators get caught out.
First, quality control is on you. White-label work going out under someone else's name with an AI tell in it — a generic phrase, a hallucinated stat — is worse than a missed deadline. Read every word. The model drafts. You're the editor of record.
Second, and people skip this: if you're using AI in client deliverables, your contract should say so. Raw AI output may carry limited or no copyright protection, which matters when you're assigning rights to a client you can't fully assign. That's not a reason to panic or hide it. It's a reason to add a clear AI clause and a sensible liability position to your agreement. Talk to someone qualified and treat it as professional hygiene, not paranoia.
Where AI does not belong in your agency
To be straight about the limits:
- The relationship. Client comms that carry emotional weight — a missed deadline, a difficult renewal, a complaint — are yours. A warm, specific human reply beats a polished AI one every time, and clients can smell the difference.
- Strategic recommendations. As above. Diagnosis and framing are the product.
- Anything you won't check. No time to read the output properly means no time to send it. Unreviewed AI in client work is a liability, not a shortcut.
- Sounding like everyone else. Your voice is part of why people hired you. Use AI to draft in your structure and tone, then make it sound like you actually wrote it. If it reads like a content mill, rewrite it.
Putting it together
The pattern across all five workflows is the same. AI collapses the blank-page problem and the formatting tax. You keep the judgment, the relationship and the final read. Onboarding briefs, status drafts, capacity maths, retainer prep, white-label formatting — that's a believable few hours a week back, spent on something a model can't do: thinking and selling.
If you'd rather not build each of these prompts and templates from scratch, the Agency Operations pillar on Skill Locker packages skills for exactly this — onboarding workflows, client status reports, capacity planning, retainer reviews, white-label report formatting. It's my product, so take that as disclosure, not a hard sell. The workflows above stand on their own whether you use it or not. Either way, the move is the same. Automate the admin, protect the judgment, and stop spending your evenings on documents that don't need a human to draft them. Only to decide what they say.