I’m several weeks into using the OpenClaw AI solution as Eliza, my head of marketing. It feels a good time to share how its gone.
There’s people using OpenClaw for all sorts of things, multiple job roles. I’ve used it for just this one for now, mainly because I’ve already got good AI automation workflows for other tasks that are better suited to my needs. As a head of marketing its been brilliant for me, or rather she, I prefer thinking of as a her, as Eliza, a real being that’s performing a real job role for my company. She’s not perfect though, all the normal AI issues are present, she can go off track, hit issues and not deal with them, she might be head of marketing but she’s also an intern at the same time. Way more skilled than me at marketing, but also prone to stupid mistakes and issues that an experienced head of marketing would just know (you can’t teach experience…)
The best bit is she’s free, or nearly free. OpenClaw is free, all it costs me is AI token usage and for now at least I’m still under the Claude AI usage radar and getting away with using their monthly fixed price subscription. That’s important, because I want opus as my AI engine due to its best in class human understanding and speaking abilities. Paying for it on an API token usage basis gets expensive and needs careful control. I’ll switch if the ban hammer swings my way, but it’ll mean work needed in setting up multiple AI providers to keep costs under control.
The biggest issue found – using a cloud server to host her on. I quickly discovered a really big flaw with this for my use case, cloud servers are really cheap (£7 a month), but they have no desktop and use known cloud provider IP addresses. The result, instantly detected as a bot by all the major social network sites. Whilst I’m not letting Eliza post on my behalf anywhere, I do need her to find where to post and what’s happening in conversations to post to. My key social network targets are Reddit (great for reach), LinkedIn (great for business audience) and X (not cracked there yet, I have terrible reach still). Solution, I bought a mac mini and moved her onto there. £500 spent, but a great investment. She now has a residential IP address and the Chrome desktop browser, so with the right tools applied can appear online as a real human. I can also sign her into throwaway LinkedIn and Reddit accounts, essential because they don’t give live API access and without it she can’t see into LinkedIn at all and can’t see anything new on Reddit (Reddit conversations die very fast). That solved that problem and overnight she became way better at finding conversations for me to post to and reach new potential clients.
I have her setup to email me daily briefings for each of my projects with suggested posts to make in relevant places and also suggested posts to make to my own social profiles. I chose whether to post them, I edit them before I do and then I use a keyboard simulator I vibe coded to actually make the post (it types the posts from the keyboard as a human would, slowly and with randomised pauses, so from the social network sites point of view its just a human typing the post). All bot detection methods defeated I think….a long as what is being posted is good…
That’s where using Claude opus is good, its great at being human. There’s been plenty of prompting Eliza needed and using her memory to set that up right so she suggests what I want posted. There’s ongoing tweaking needed to keep her on track, but overall I’m really happy with her work. Its not that I couldn’t do it myself, but I’m busy, very time poor, I never find the time to be able to. Having her on the team, my socials are doing what they need to be for marketing, reaching new people with helpful and informative content. Its essential to me, because AI has pretty much killed my previous marketing through all the resource sites I’ve created over the last 10 years, their traffic fell off a cliff about 6 moths ago when Google started answering queries directly.
On top of this her daily briefings give me evergreen content ideas to post on the news sections of my websites, to build my SEO (another area I’d never found the time to do properly). I’ve also set her up to regularly check social network sites throughout the working day for new highly relevant posts I should comment on and email me immediately with a suggest comment. She’s also on the end of Telegram messenger so I can instruct her whenever I think of anything new I want her to action or tweak.
Another really big benefit of having been through this process is what I’ve learnt about marketing along the way. Some of what she suggests I’d never have come up with, I’m not always sure its all that interesting. But some of those posts get picked up by the algorithms and get way more reach than anything I’d have come up with. Each time it happens I get a bit smarter, I learn a bit more about marketing and whats needed from it today.
She’s not perfect, she’s got lots of complexity at the tech setup and maintenance level and keeping her bot detector friendly. She needs new guidance every now and then. But now she’s setup she’s an essential member of my team, costing me ~£150 a month to run and <0.5hour a day, outside of the admin side of her.
Do I like that AI is taking over socials? Nope, its a real problem. But I’m a realist. AI has killed my previous marketing approach so the only choice is to evolve and move with the times on it.
Btw, if you’re interested in what OpenClaw (or any other AI) could do for your business, this is exactly the sort of thing I do for clients. Its part of what I’m having her work on marketing for me. Through my company I design software for clients, build AI systems, build cloud platforms, integrate into businesses, create perfect user experiences, design great products. Get in touch if you’d like to explore this area for your business!
