The short version: I run a one-person consultancy with clients in the UK, US, and Israel, and AI tools have cut my working hours by roughly a third without cutting my output. This post is the exact stack I use every single day in 2026, what I use each tool for, how much it costs me, and the things most "best AI tools" roundups will never tell you.
I want to be upfront about something before we start. I am not an unbiased observer here. I have tested dozens of AI tools over the past three years, I have spent real money on them, I have been disappointed by most of them, and I have found a small handful that I would miss immediately if they disappeared tomorrow. Those are the ones I am writing about today.
I am also writing this from a specific context. I rebuilt my business almost from scratch after five difficult years. Money mattered. Time mattered. I could not afford tools that were impressive in demos but useless on a Tuesday morning when I had four client deliverables due. Every tool in this list earned its place by surviving that filter.
Why I stopped rotating tools every month
For about eighteen months I was guilty of what I now call "AI tourism." I would read a newsletter, see a new tool, sign up, play with it for a week, get distracted by the next one. I had accounts on at least thirty platforms. I was producing nothing useful and spending a lot of time feeling like I was being productive.
The shift happened when I did something embarrassingly simple. I listed every tool I had paid for in the previous six months and asked myself: did I open this more than five times? For most of them, the answer was no. I cancelled fourteen subscriptions in one afternoon. You can read the full breakdown of what stayed and what went in my post on the marketing tools I pay for in 2026 and the ones I cancelled.
What I kept was a core stack of tools I open within the first hour of every working day. That is the only meaningful definition of "daily use" I trust. Not "I use it sometimes." Not "it is in my bookmarks." Every single day.
The daily stack: what it is and what it costs
Before I go tool by tool, here is the honest total. My AI tool spend right now is approximately £180 per month. That sounds like a lot until you compare it to the hours it replaces. At my consulting day rate, if these tools save me ten billable hours a month (conservative estimate) the return is enormous. But I am not going to pretend the costs do not matter, especially to someone earlier in their business journey than I am.
ChatGPT: the one I open first
I use ChatGPT on the paid plan and I open it before I open my email. That tells you something.
What I use it for is not what most people expect. I do not use it to write blog posts for me. I use it to think. My morning routine includes what I call a "brain dump session" where I paste in whatever is cluttering my head, whether that is a client problem, a proposal I am struggling to frame, or a piece of feedback that stung, and I ask ChatGPT to help me find the actual question underneath the mess. Nine times out of ten it surfaces something useful.
I also use it for first-draft structures. Not the content. The bones. If I need to write a strategy document for a client, I will describe the situation, ask for three possible structural approaches, pick the one that fits, then write the actual content myself. That step alone saves me forty minutes on a complex document.
For client work specifically, the single most valuable thing I have done is build detailed custom instructions that tell ChatGPT exactly who I am, what kind of clients I serve, my tone of voice, and what I never want it to assume. If you have not done this yet, stop reading and do it now. It transforms the quality of outputs immediately. I wrote a full guide on how I use AI to write briefs that get good work back which covers this in detail.
Cost: £20 per month for the Plus plan. Non-negotiable for me.
Claude: the one that thinks longer thoughts
I use Claude for different things than ChatGPT and I think this distinction matters more than most people realise.
Claude is where I go when the task requires holding a lot of context at once. Long documents. Complex client situations with multiple moving parts. When I am reviewing a 40-page strategy deck someone has sent me and I need a thoughtful synthesis rather than a quick answer, Claude handles that better than anything else I have tested.
I also use Claude for tone-sensitive writing. Anything where I need the output to sound human and considered rather than "AI confident" (you know that voice, slightly too polished, slightly too certain). Claude has a more careful, almost cautious quality that suits some contexts very well.
One specific example. A client last year was dealing with a difficult internal stakeholder situation. They needed a communication strategy that was honest but not inflammatory. I put the whole situation into Claude, including emails, the background, the politics, the desired outcome, and asked it to help me structure an approach. What came back was not a finished document. It was a framework with twelve specific points I had not thought of. That session probably saved the client relationship.
Cost: approximately £18 per month on the Pro plan.
Perplexity: the one most people have not tried yet
This is the tool I recommend most enthusiastically to people who are still using Google for research and finding it increasingly frustrating.
Perplexity is an AI search tool that gives you sourced answers. You ask a question, it finds recent information, it cites where it got it from, and you can follow up. It is significantly better than standard search for anything research-intensive because it synthesises rather than lists.
I use it every single day for competitive research, market context, and client industry briefs. When I take on a new client in an industry I do not know deeply, the first thing I do is spend thirty minutes in Perplexity building a working understanding. I ask layered questions: who are the main players, what are the current pressure points, what has changed in the last twelve months, what are people in this space complaining about. I leave that session with a solid foundation that would have taken me two to three hours of old-fashioned research to build.
The honest caveat: it still hallucinates sometimes. I always verify anything specific, any statistic, any claim, before I repeat it to a client. But compared to a standard search engine, the signal to noise ratio is dramatically better for professional research tasks.
Cost: around £16 per month for the Pro plan. There is a free tier but the daily query limit is frustrating if you use it heavily.
A design AI I use that is not what you would expect
I want to be honest here. I am not a designer. I never have been. For most of my career I either hired designers or used templates and hoped for the best.
What AI has done for me in the design space is not replace designers for complex work. It is handle the hundred small visual tasks that used to take me out of flow completely. Resizing an image to the right dimensions. Creating a simple social graphic that matches my brand. Generating a visual to go with a blog post. That kind of thing.
I use Canva's AI features for most of this, specifically the Magic Design and background removal tools, and I use Adobe Express for certain client-facing assets where the output quality needs to be slightly higher. I wrote a proper head-to-head on this because the difference between them is more nuanced than most people realise. If you are trying to decide between the two, read my honest verdict on Canva vs Adobe Express for small business design before spending money on either.
I also use AI image generation (I switch between a couple of tools depending on the job) for featured images on this blog and for client presentations. The time saving is real. What used to take me forty-five minutes hunting through stock photo sites now takes about eight minutes.
The email and writing tool I use for everything I publish
I write a lot. Blog posts, client reports, newsletters, social content. My actual writing process has not changed much but my editing process has changed completely.
After I write a first draft, I use an AI editing pass to catch specific things: sentences that are too long, repeated words, places where I have been vague when I should be specific, and places where I have softened an opinion I should have stated plainly. I use this as a checklist, not as an instruction to rewrite. The distinction matters enormously. The AI does not rewrite my work. It flags issues. I decide whether to fix them and how.
For newsletters specifically, I run every draft through a prompt I have built over about eighteen months that checks for five things: clarity of the main point, whether the opening would make someone keep reading, whether there are any claims I have not backed up, whether the call to action is specific, and whether the tone matches my brand voice. That prompt has made my newsletters better. I see it in the reply rate.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
The honest thing most AI tool articles will not tell you
Here it is. The thing I almost never see written plainly.
AI tools make bad thinking faster. That is not a reason to avoid them. But it is a reason to understand what is happening when you use them.
If you go into a ChatGPT session with a fuzzy question and vague context, you will get a fluent-sounding answer that is also fuzzy and vague. The tool will not save you. The tool will produce faster what you brought in. The quality of your output is almost entirely determined by the quality of your input. I have watched people use these tools for months and conclude they are disappointing, when what was disappointing was the prompts, not the tool.
The consultants and creators I know who get extraordinary results from AI have one thing in common. They think carefully before they type. They have a clear question. They provide specific context. They tell the tool what they want the output to do, not just what they want it to say. That takes practice and it takes a willingness to write longer, more specific prompts than feels natural at first.
I spent roughly three months iterating on my prompts for my most common tasks before the outputs were good enough to use with minimal editing. Most people give up before that. I am not saying that to be discouraging. I am saying it because if you are two months in and still frustrated, that is completely normal and you are not doing it wrong.
What I have tried and rejected
I think it is worth naming a few categories of tools that did not survive my filter, because the hype around them is significant.
AI social media scheduling tools that claim to generate and post your content automatically: I tried three of them. The content quality was uniformly poor and it sounded nothing like me. Scheduling is fine. Auto-generating the content is not something I am willing to outsource to a tool that does not know my voice, my audience, or what I said last week.
AI SEO tools with "one-click optimisation": I tested two of these, running the same content with and without their suggestions over three months. The results were indistinguishable. I kept using my own process. If you want to know what I think is worth spending money on in the marketing tools space, and what is not, I covered the free alternatives in my post on the best free alternatives to expensive marketing tools.
AI chatbots for customer service on small websites: the setup time was enormous, the maintenance ongoing, and the conversations I reviewed were often misleading. For a business at my size and stage, not worth it.
How my AI tool use connects to client work
A lot of what I do professionally is help businesses think about where AI fits into their marketing and operations. And something I have noticed is that the businesses getting the most value from AI are not the ones with the biggest budgets or the most sophisticated tools. They are the ones who have identified two or three specific, high-frequency tasks where AI helps and then built habits around those tasks.
For one client, a small e-commerce business, the highest-value AI application we found was not content generation or customer service. It was using AI to analyse their customer reviews at scale and surface recurring themes they had never noticed before. That insight changed their product descriptions and their returns rate dropped by 11% in two months. One specific use case. Enormous impact.
For another client running a services business, the win was using AI to create first drafts of client proposals so their team could spend less time on the blank-page problem and more time on the bespoke elements that win the work. Proposal turnaround time went from five days to two days. That directly affected how many proposals they could send per month.
The point is not that my stack is the right stack for everyone. It is that specificity beats comprehensiveness every time. If you are just getting started and trying to work out how AI might help your business, I would also point you toward my thoughts on what is working for small business marketing in 2026 more broadly, because AI tools do not exist in isolation from the rest of your marketing strategy.
My actual daily workflow with these tools
This is step by step, because I think seeing the sequence matters.
- 7:30am: Open ChatGPT. Brain dump anything that is already in my head. Clear the clutter before I look at anything else.
- 8:00am: Check email. Use Perplexity to research anything a client has raised that I need context on before I respond.
- 9:00am to 12:00pm: Client work. ChatGPT open in a tab. Claude open in another. I switch based on the task: quick structured thinking goes to ChatGPT, long-form analysis or tone-sensitive writing goes to Claude.
- 12:00pm: Any visual assets I need get handled in Canva with AI features during lunch so they are not interrupting my deep work time.
- 2:00pm to 4:00pm: Writing for my own business. Blog posts, newsletter, social content. I write first, then run the AI editing pass, then review the flags and decide what to change.
- 4:00pm: Admin, invoicing, planning. I use ChatGPT occasionally here for things like drafting difficult emails where I want to think through the framing before I write.
Total time in AI tools: probably ninety minutes to two hours per day across all of them, but woven into the work rather than separate from it. That is the key point. These are not things I do instead of working. They are part of how I work.
What I am watching in 2026
A few things are developing that I think will change this stack within the next twelve months.
Agents. AI that can take a task, break it into steps, execute those steps across different tools, and come back with a result rather than just a piece of text. I have tested early versions of this and they are impressive but not reliable enough yet for client-facing work. I am watching closely.
Voice input. I have started dictating more of my AI prompts rather than typing them and the quality of what I produce has gone up because I am less tempted to be brief when I am speaking. If you have not tried voice input for your AI interactions, experiment with it.
Memory and personalisation. The tools that build a genuine understanding of your preferences, your clients, your projects over time will be significantly more valuable than tools that start from zero every session. Some of this is already available. It will improve quickly.
I will keep writing about all of this as it develops. That is what this blog is for: the specific, honest, lived experience of building a business in public in a strange moment for the industry.
Frequently asked questions
Do you need multiple AI tools or can one tool do everything?
One tool cannot do everything well in 2026. ChatGPT, Claude, and Perplexity each have genuine strengths in different areas: fast structured thinking, long-context analysis, and sourced research respectively. Running two or three tools costs less than £60 per month combined, and the quality difference between using the right tool for a task versus the wrong one is significant enough to justify it.
How long does it take to get useful results from AI tools?
Based on my own experience and watching clients: expect two to three months before your prompts are good enough to produce outputs that need minimal editing. Most people give up in the first four weeks when the quality is still inconsistent. The quality improves as your prompts improve, not as the tool improves. That is the honest timeline.
Are there AI tools that are not worth the money?
Yes. AI social tools that auto-generate and post content without human review, one-click SEO optimisation tools (the results in my three-month test were indistinguishable from doing nothing), and AI customer service chatbots for small businesses where the setup and maintenance time vastly outweighs the benefit. Save your money on those categories.
What is the single most impactful change I can make to get better results from AI tools I already have?
Write longer, more specific prompts. Include who you are, what the output is for, who will read it, what you want them to do or feel after reading it, and what you do not want the output to include. Most people write three-sentence prompts and wonder why the output is generic. Prompts that take two minutes to write produce outputs that are dramatically better than prompts that take thirty seconds.
Related reading: No Experience Jobs Work From Home: The Honest Guide to Getting Started in 2026 and Earning Passive Income: The Honest Guide From Someone Who Has Done It, Lost It, and Built It Back.
