May 2025 wrap-up
What have I been doing and learning this month? This blog post triples as my notes, a status update, and a way to share the things I found interesting this month.
Highlights
Getting to play with Kapa (RAG chatbot for your docs sites and communities, plus a bunch of other useful stuff). The more I explore it, the more impressed I am.
Strong feature set, good docs, and pretty straightforward to get up and running (especially considering they don't officially offer self-serve yet). I get the feeling this is a team of people who really want to help users find information, and also want to help companies understand their users and improve their docs.
This seems like an actually good use of AI.
Writing
The main work for this month and next is Documentation for Bolt: the overhaul is in full swing, along with some new content. This contract is giving me a chance to try out Mintlify, which I've been wanting to do for ages. It's an interesting tool: it solves one of the big problems with docs like code (how to collaborate with people who can't use git), but the price seems to be some fairly tight limits on customisation, and a relatively limited feature set. However, it's a young product, so these may well be problems that get fixed with time.
I made a start on guidance for adding Kapa AI search to a Material for MkDocs site
Building
Finally sorted out my own self-hosted n8n instance on DigitalOcean. I followed n8n's docs Hosting n8n on DigitalOcean, with the addition of the environment variables to Isolate n8n. This is something I really should have done while working there, but somehow never got round to.
Watching
The Marketing Meetup: How to have impactful ideas - Pieter-Paul von Weiler and Matt Davies
A talk about idea evaluation: how to know if you've got a good idea? Includes a bunch of stats on views from a survey of marketers and agencies (yay for data!).
Strategy -> Brief -> Creative evaluation
The poor state of ideas
- Most creative work doesn't stand out.
- The quality of creating work is not progressing.
- There's a lack of pride in the creative work that's made.
- Evaluating ideas is hard.
Interesting to note both marketers and agencies back this, but agencies are a lot more strongly pessimistic than marketers.
What's holding better ideas back?
- Lack of training.
- Feedback provided is often unclear and inadequate.
- Approval processes aren't working well: inconsistent, slow, subjective, painful.
- Creativity dies in committees. Trying to accommodate everyone makes things worse. The person who signs off the brief should sign off the idea.
- Briefs aren't being used when ideas get evaluated. When evaluating, you should quickly recap the brief to frame the evaluation.
- Missing criteria for evaluation.
- Personal opinion is shaping feedback and decisions. You don't have to like an idea for it to be right for your brand and target audience.
The impact of poor assessment practices:
- The rework rate is high and time consuming.
- Agencies lack trust in marketers.
- Expensive!
Tips to get to better ideas
- Empower agencies to do their best work (marketers think they're doing this, agencies disagree).
- Write better briefs.
- Refer back to the brief when evaluating an idea.
- Use these criteria to evaluate an idea:
- Is it engaging for the target audience?
- Is it on brief?
- Will it earn people's attention?
- Is it well-branded?
The Marketing Meetup: How marketers can use AI to be more efficient - Nick Crawford
I had quite a lot of opinions throughout this talk, which I've noted. I end up sounding quite critical of this talk, which isn't fair. It was a clear, well-structured presentation, and introduced me to a couple of tools I'm interested in trying. However, I do think it glossed over the amount of monitoring AI needs to be safe.
Intro:
- AI can save us time.
- Issues to be aware of: AI bias, hallucination, sustainability.
- My note: love that he put these caveats up front.
- AI won't take your job, but it will take our pressure.
Five use cases:
- Presentations: turning ideas into polished decks.
- Research: helping us absorb large amounts of content.
- Data analysis
- Idea generation: get first drafts.
- Everyday stuff: have a play, see what you can use it for in personal life.
Presentations
- Use ChatGPT to give us the outline: feed in material, ask for a presentation based on it.
- My note: this felt really risky, or (if done safely) not a time saver. He asked ChatGPT to do data analysis, to save time going through the data and finding trends. But it's perfectly possible for the AI to make things up here, so you'd still have to do the review work.
- Gamma to create the visual presentation.
- My note: this bit looked more useful.
- Napkin for graphics.
Absorbing information
Use NotebookLM: add many types of sources, get summaries, generate FAQs . . . and autogenerate podcasts.
My note: love the summaries. Gut instinct hate the idea of autogenerated podcasts as anything more than an audio summary for people who prefer that format.
Data analysis
Feed data into ChatGPT, ask for a summary and visuals.
My note: again, this makes me nervous and I'd question the wisdom of relying on it.
Idea generation
For this one he used Claude - reckons it's more human-sounding.
Example: ask it to suggest three campaign hooks for a particular target audience.
My note: I still stick by my Thinking first, AI second approach.
Everyday stuff: ways to play with AI
Ask for:
- Meal suggestions based on a photo of your fridge.
- Use it as a timeblocking assistant.
- Set up a 7-day habit challenge.
- Plan a weekend break.
My note: I feel I'm being overly negative at this point, but none of these appeal. Why do I need AI to suggest cooking ideas, I'm perfectly capable of glancing at the food I have in and being creating. Part of the value of planning is doing the planning (you're still going to have to review what the AI suggests). And planning breaks is part of the fun! In short, these suggestions feel like they're asking AI to do things that my brain actually enjoys or benefits from doing.
How to identify places to use AI
Ask yourself:
- What tasks feel repetitive or dull?
- Where do you need creativity but lack time?
- What would you delegate to an assistant?
- What AI use surprised you today?
- What would '10% more efficient' look like?
- Where are you 'data rich but time poor'?
- What hesitations about AI do you have?
- What fun AI project could you try?
The Marketing Meetup: How to grow an agency in 2025 - Stephen Kenwright
This was in Q&A format, so notes are inevitably a little messy.
Two challenges:
- Not growing enough
- Too much growth
Agencies are susceptible to industry-wide challenges. At the moment: lack of economic growth, chaotic business environment, post-pandemic aftermath, and AI disruption.
Two modes:
- Do a thing, repeatedly, and sell it as a service. This is more vulnerable to disruption.
- Develop custom solutions to specific client problems. This one won't scale as large, but you can charge more, and custom strategic work is less vulnerable to AI disruption.
Need to understand who you're trying to serve. Not just industry verticals. Think about which brands have the same challenges as each other.
Your competition isn't just other agencies, but also in-house solutions.
Getting leads:
- Pick a channel and commit. Invest where you're comfortable. You'll need time - months of consistent LinkedIn posting, for instance. You can't really outsource this: the owner needs to be out there driving inbound leads.
- If you're busy, this is a good time to do marketing! Look at what you're busy with, and decide if it's the right thing to be busy with. Make sure to keep doing your marketing and generating opportunities. You want enough opportunities that you can walk away or turn down projects that aren't a fit.
Client relationships: Hiring freezes and cutbacks can actually be an opportunity for agencies. Requires the agency to be very clear about what they can do.
Hiring for agency:
- Typically follow a certain pattern. First, hire someone who does some of the work (you continue to be the person who leads and talks to clients). Probably someone fairly junior. Eventually, you need to hire someone to organise/coordinate (when you reach the point you can't oversee everything). You eventually become mainly client-facing, and end up not 'doing the thing' as much any more. So ask yourself before you start: what do you actually want? If you want to be hands-on in the problems, you likely want to stay smaller.
- First hire needs to be able to do some of the work, and be someone you are happy to work with and train.
Identifiable challenges at different stages:
- ~8 people: can't manage everyone, so hire a people manager.
- ~16 people: someone is needed to manage your people managers. You now have people you can't talk to, and need to start managing through managers.
- ~50-60 people: you don't know everyone's name any more.
You parachute into my one-person B2B branding/website agency for 90 days. What are the first three moves you’d make to flip it into a scalable & growing agency? (General priorites)
- Cash position: is there runway?
- Pipeline: what opportunities are coming in? If plenty, you can make lots of decisions (be choosy about work etc.). If there isn't pipeline, the top priority is to get that in place.
- Get clear about who you're for.
ScoreApp: Learn the 5 Secrets to bring the right clients - Martin Huntbach
Webinar recording on scoreapp.com
This webinar is relevant if:
- You're a business owner with something to sell.
- You care about working with good people.
- You're committed to sharing your knowledge.
This was a helpful webinar, but quite hard to note-take live (it included things like charts to help you benchmark your current client leads, and interactive elements). It discussed an approach to building a steady pipeline of good-quality clients: have a tool/process that gives personalised guidance (but automated).
Personalisation (at scale) is the best way to attract high-quality clients consistently.
When Martin tried this, he found:
- Increased his authority.
- Leads 10x'd
- He had to increase his prices to reduce demand.
People will follow you for value, opt-in for resources, and hire you if you consistenlty solve their problem.
Listen first, advise second.
You need a filtering process to identify clients that are a good fit. 81% of entrepeneurs have had a difficult client in the last year.
My note on the elite client funnel: this risks feeling less personal to me. I'd be irritated if I requested a call with a consultant, then got pushed into a content funnel. Maybe it works for agencies doing bulk, getting a lot of enquiries, and wanting to filter people out? But I'd be wary of applying it to my own business.
Worth noting the webinar is by ScoreApp, who sell a tool to help attract clients. So a pinch of salt on some of the impact claims. But it is a useful perspective, and I'd certainly love to build a quick docs audit tool.
Your Product Needs Words