The AI Tools I Use to Run My Marketing Business (And What Each One Replaced)

The short version: I run my marketing business on about eight AI tools that between them replaced a research assistant, a copywriter, a transcriptionist, a social media manager, and roughly half of what I used to pay freelancers each month. None of them replaced judgment, and the tool I use least often is the one I get asked about most.

Why I'm even writing this down

Someone asked me at a conference in Manchester last month what my "AI stack" was, and I realised I'd never written it out. I've been rebuilding this business for five years now, mostly in public, and the tools have changed more than the strategy has. So here's what's on my laptop, what it cost to swap, and what I stopped paying someone else to do.

I'm not going to pretend every tool changed my life. Some of them just made an annoying job faster. A couple changed how many clients I can serve without hiring. I'll tell you which is which.

ChatGPT and Claude: what used to be a junior copywriter's job

I use both, for different reasons. ChatGPT (the paid Plus tier, £20 a month) does my first-draft social captions, email subject line testing, and quick rewrites when a client sends me three paragraphs of jargon and asks me to "make it sound human." Claude is what I use for anything longer: proposal drafts, blog outlines, the first pass of a report.

Here's the uncomfortable bit nobody likes saying out loud: this used to be a job. I paid a freelance junior copywriter £25 an hour for exactly this work for about three years, roughly £400 a month for first drafts I then edited. That person doesn't get that work from me anymore. I don't feel great about it, and I'm not going to pretend I do, but I also can't run a business paying for a first draft I can produce myself in four minutes and then improve in fifteen. That's the trade-off nobody in the "AI made me so productive" posts wants to mention. Productivity gains for me are sometimes someone else's lost invoice.

What it replaced specifically: the £400 a month junior copywriting retainer, gone. What it didn't replace: the editing eye that knows a client's actual voice, which is still entirely me, an hour or two a week per client.

Perplexity: what used to be two hours of Googling

I use Perplexity for research the way I used to use a search engine and eight open tabs. Client wants to know what their competitor's pricing model looks like, or what's changed in a niche regulation, or who's currently ranking for a term. Perplexity gives me sourced answers in under a minute instead of the old routine of Googling, opening five results, skimming, and cross-checking.

It replaced roughly two hours a week of manual research per active client. At my rate, that's real money back in the week, not just a nice-to-have. If you want a fuller comparison of what different AI tools are good and bad at for this kind of work, I wrote a longer breakdown of which AI tools are best for marketing research, including where each one lets you down, because they all do, somewhere.

What it didn't replace: primary research. If a client needs to know what their own customers think, no AI research tool does that. You still need surveys, interviews, and the free tools I've used for years to understand what an audience cares about, which is a different job entirely from summarising the internet.

Otter.ai: what used to be an hour of my evening

Every client call gets recorded and transcribed through Otter. This one is boring to explain but it's probably saved me more actual hours than anything else on this list. I used to type up my own notes after calls, sometimes at 9pm with a glass of wine, trying to remember what someone said forty minutes earlier. Now I get a searchable transcript with action items flagged before I've even closed my laptop.

What it replaced: roughly five to seven hours a week of my own note-taking and follow-up email drafting. That's not a small thing when you're running a one-person consultancy alongside speaking engagements and this blog.

Descript: what used to be a video editor on retainer

I do a fair bit of video, for LinkedIn, for client training, for the odd webinar recording that needs turning into something usable. I used to pay a freelance editor around £150 to £250 per finished video, depending on length. Descript lets me cut a video by deleting text in a transcript, which sounds like a gimmick until you've done it and realised how much faster it is than scrubbing a timeline.

What it replaced: the editor, for anything under about ten minutes. I still send longer, more polished pieces (like anything going on a proper sales page) to a human editor, because Descript's output is good, not exceptional, and clients can tell the difference on anything high-stakes.

Buffer plus a scheduling assistant: what used to be a social media manager

This is the one that surprises people. For years I had someone part-time managing scheduling and light community replies across platforms. Now I plan a week of content in about ninety minutes using ChatGPT for variations and Buffer to queue it, and I check notifications myself twice a day. It's not glamorous, but it's honest: for a business my size, a dedicated social media manager was overkill once the drafting itself got faster.

If you're managing anything on X, the platform has got fiddlier over the past two years, verification changes, video length limits, algorithm shifts, and I keep a running reference for exactly that kind of thing in my Twitter/X help guide, because half the questions I get from clients are "why did my video get cut off" rather than anything strategic.

What it replaced: roughly £600 a month in part-time social media management. What it didn't replace: strategy, which platforms deserve attention this quarter, that's still a conversation I have with every client myself.

Fathom: what used to be a shared notes doc nobody updated

This overlaps a little with Otter but Fathom's specific job is meeting summaries for internal team calls and client check-ins, with action items auto-assigned. Small thing, but it replaced the shared Google Doc that three people were supposed to update after every call and one person did. Now nobody has to remember; it just lands in Slack.

What I use for customer service now, and what surprised me

I don't run a huge support inbox, but between the blog, the newsletter, and a couple of client projects, questions come in constantly. I started testing conversational AI for first-line responses last year, and the thing that changed my mind wasn't the speed, it was the tone. Done badly, this feels like a wall. Done well, it catches people before they get frustrated. I wrote about this, including where it helps a small team rather than just looking impressive in a demo, in my piece on how AI customer service tools help small teams handle more without burning out.

What it replaced: nothing I was paying for before, honestly, because before this I just didn't answer fast enough and lost a few people to slow replies. This one's a genuine addition rather than a swap.

A step-by-step of what my actual week looks like

To make this concrete rather than a list of logos, here's Monday for me right now:

• 8am: Otter transcript from Friday's client call, reviewed in ten minutes, action items copied into my task board
• 8.30am: Perplexity for one competitor check a client asked about, twelve minutes, sourced and screenshotted
• 9am: Claude drafts a report skeleton from bullet points I dictate into my phone on the school run
• 10am: I edit that draft, this is the part that's still entirely mine, about forty minutes
• 11am: ChatGPT generates five social post variations from the report's key point, I pick two and tweak
• 11.20am: Buffer queues the week's content, scheduled across platforms
• Afternoon: actual client calls and the work that no tool touches, strategy, pricing conversations, saying no to scope creep

I keep the whole week visible on a simple board rather than trying to hold it in my head, which sounds basic but stopped things falling through the cracks. If you've never tried a visual system for this, I wrote about how kanban style boards help you see task progress at a glance, and it's one of the least exciting changes I've made that's had the most day-to-day effect.

What none of this replaced, and why that matters

Here's the part I think most people skip because it's less flattering: these tools cut my monthly outgoing from roughly £3,200 in freelance and part-time support down to about £310 in software subscriptions, and I want to be honest that the difference didn't all go to profit. Some of it went to redoing work that AI got confidently wrong, a fabricated statistic in a draft, a transcript that missed a client's actual instruction, a social post that used a phrase a client hates. The saving is real but it's not the full number on the spreadsheet.

What no tool on this list has replaced: the fifteen years of knowing which client is about to churn before they say anything, the judgment call on whether a campaign needs boldness or restraint, and the actual relationship that gets a client to pick up the phone when something's gone wrong. If you're weighing up whether to build a stack like this yourself or bring someone in to set it up, that's a fair question, and it's worth knowing roughly what an AI consultant costs before you decide either way, because DIY isn't always cheaper once you count the hours you'll spend fixing your own mistakes.

My honest advice after five years of swapping tools in and out: pick the one that removes the task you hate most first, not the one everyone's posting about. I didn't start with ChatGPT because it was trendy. I started with Otter because I was sick of typing up my own notes at 9pm.

See also The AI Tools I Use to Run My Marketing Business and What Each Replaced, which picks up where this leaves off.

Related: AI write for us.

Related: internet marketing: guidelines and how to pitch.

Frequently asked questions

What's the single AI tool that saved me the most money?

Claude and ChatGPT together, because they replaced a standing £400 a month copywriting retainer for first drafts, though I still spend an hour or two editing every draft myself before it goes anywhere near a client.

Do AI tools replace the need for a marketing team?

They replace specific repeatable tasks, drafting, transcription, scheduling, first-pass research, not judgment, strategy, or client relationships. I run a smaller team than I used to precisely because those repeatable tasks got cheaper, but the strategic work is still entirely human.

How much should a small business budget for AI tools each month?

I spend around £310 a month across eight tools, which is a reasonable starting range for a solo consultant or small team. Most tools have free or low-cost tiers, so I'd trial two or three at the entry level before committing to paid plans for all of them at once.

Which AI tool did I try and drop?

I tested a couple of all-in-one "AI marketing platforms" that promised to do everything from ads to email to social in one dashboard, and dropped both within a month because each individual function was worse than the specialist tool I already used. Jack of all trades software is still the exception here, not the rule.

Where to check the details

• Anthropic's own documentation
• the NIST AI Risk Management Framework