The AI Toolkit: Which AI Tools to Use in Your Business, and When

This was session two of our AI Week, and the promise was narrow: by the end you should know how to spot a good tool, meet a few of my favourites you might not have tried, and understand how to use specialist tools in the right moment. If you'd rather watch than read, the full session is embedded further down. Otherwise, here's the toolkit.

Nobody feels on top of AI, and that's fine

The last two years have generated as much information as every year before them combined. Every tool on the market has an infinite budget and the smartest people in the world working on it, and they all look like the best thing ever built. I keep calling it the AI arms race: they copy each other's features the moment one lands. So you aren't behind: everyone is behind. The goal isn't to catch the whole market; it's to build a small toolkit that does your actual work.

AI tools are tools: hire them, don't collect them

A lot of people think you have to catch them all: pay for hundreds of subscriptions, own every tool on the market. You don't. You can get a huge amount done with a small set of everyday tools. The shift is to stop collecting and start hiring, almost like mercenaries: bring a tool in as and when you need it, and only if it earns its place. Before anything joins your essential stack, make it answer four questions.

1. What recurring business task does it actually perform?
2. Who is responsible for using and improving it? You want a champion inside the business, not a tool nobody owns.
3. What output does it produce: an asset, a decision, an automation, a customer result?
4. What does success look like, so you can tell at 30 days whether it's worth keeping?

If you can't name the job, the owner or the outcome, don't adopt it. Most of these tools run generous free plans anyway, so you can explore far deeper than you probably need before paying a penny. Used well, a tool amplifies the team you have: an AI-augmented team gets three to five times the output of a standard one. I wrote up that whole argument in why AI isn't optional anymore.

The six buckets: which tool for which job

Here's how I sort every AI tool and agent in my head. Six buckets, defined by the job, not the brand. Feel free to adjust them to fit your own head, but for most people one tool per bucket is more than enough.

Figuring out which tool excels at which task took me months, a lot of money, and plenty of breaking things and rebuilding them, so I've done that part for you. Even if you paid for every tool I mention, with all the overlap you're looking at two to three hundred pounds a month: set that against a fraction of one salary and the maths isn't close.

Bucket one, everyday AI: ChatGPT, Gemini and Claude

These are the tools you live in: flexible across many use cases, fast, and conversational enough that you get things done just by talking to them. For most people this bucket is the whole of their AI knowledge, and it comes down to three names.

ChatGPT is the one that made the public realise AI was a thing. It's on par with the others for raw capability, but its superpower is obedience: it does exactly what you tell it. Give Gemini or Claude a list or a workflow and they often forget a step or decide they know better. If you need instructions followed to the letter, ChatGPT is your go-to, and its deep research function is strong too.

Gemini is Google's model, and where ChatGPT wins on obedience, Gemini wins on modality. It's the only one that can natively watch a video, listen to audio and make sense of a massive file. I handed it a messy three-minute Loom of me showing my team how to tidy their Slack channels, and it turned it into a clean SOP within seconds. Its reasoning lags the others, but the real moat is Workspace: signed in as your Workspace account it can query your entire Google environment, and it's bundled into most Google business plans at around £10 a month.

Claude, by Anthropic, is in my opinion the best everyday model right now. It writes the highest-quality first drafts of anything and is far ahead on copywriting and code. It'll draw a diagram straight in the chat, and its artefacts are custom little tools that run inside the chat with no deploying or hosting: I asked it to build a live dashboard for my weekly KPIs and it produced one that updates in real time as I type. That habit of ending your prompt with "ask me some clarifying questions to get the best output" makes almost any output ten times better.

Colin Cowan, CEO and founder of ETTA Energy, on his key takeaway from AI Week: "For me, the key takeaways have been how to use AI as a toolbox. As Kai said, you wouldn't use a hammer for what a screwdriver does. I've become a pro user of Claude, using it to create presentations in my company colours and profile."

You don't need all three. You need to know which one fits the job in front of you.

Bucket two, search and synthesis: Perplexity, NotebookLM and Poppy

These tools find, understand and combine information from many sources in minutes. One thing worth knowing: they're tools, not foundational models. A tool like Perplexity takes an underlying model's intelligence and moulds it to be very good at one job, which is exactly why you only reach for it now and then.

Perplexity is my pocket research assistant: the one I use to fact-check a statistic or start research on a project. Its trick is the Google-style site operator: ask about the best supplements for sleep with "site: Reddit" on the end and it only pulls from Reddit, which is what you want when you're researching how real people talk about a product. It synthesises the answer and sources every claim. NotebookLM is your internal training architect: you give it your own sources and it only answers from those, so it can't hallucinate: feed it your handbook, policies and procedures as markdown and it turns them into a chatbot, flashcards, even a podcast. Poppy is an interactive mind map: attach your brand, tone of voice, videos and competitor content, pull several models against them at once, and work that used to take days happens visually in minutes.

Bucket three, the creative studio

This bucket turns rough ideas into polished, shareable assets: design, media, presentations, video. I don't use these to replace my designer or editor: AI doesn't understand taste. I use them to hand my team a real concept in minutes instead of a vague voice note. Nano Banana Pro is a strong image generator; Google Flow does the same for video; Claude Design builds designer-ready briefs from your brand book and writes real code to render animated HTML mockups. In every case the output still reads a little too AI to my eye, but it lets me articulate exactly what I want and hand my designer a starting point he can make ten times better within the hour.

Bucket four, autonomous agents

These don't just assist, they act on your behalf. Give one a goal and it'll plan, execute and iterate across a multi-step task with little or no input. Two lead the field: Manus and Claude Code. Manus is the one to start with: quick wins straight off the shelf; Claude Code is more powerful and scalable but needs more setup. Manus is like a new employee for twenty dollars a month: I talked it through a client site build line by line and it produced a spreadsheet assigning each task to the right department, roughly 90% done, turning a few hours of admin into under an hour.

Bucket five, smart assets

These turn everyday outputs into dynamic, reusable assets rather than static files. Google Workspace is the central operating system for us: email, storage, meetings, chat, tasks, plus Gemini and NotebookLM wired in, for around £10 a month per user. Gamma turns a meeting transcript into a branded follow-up deck: summarise the transcript into slide-ready points, feed it to Gamma with your theme locked in, and send a prospect their pain points, your offer and the next steps within fifteen minutes of the call. Whisper Flow is voice-to-text across all your devices: it's free and very good, and it changed how I work because the friction of giving context disappears when you can just talk.

Bucket six, automation and orchestration

This bucket connects your apps, moves data and runs repeatable workflows on conditional logic: when this happens, do that. Zapier, Make and n8n are roughly beginner, intermediate and advanced. One of my n8n workflows saves me 16 hours a week: every six hours it checks which client ad accounts are live across Google and Meta, pulls their stats, and updates or pauses each row in Airtable.

Map the process before you build it, which is the whole point of how to map your AI and automation workflows.

How to actually use them: be the orchestrator

Owning six tools isn't the skill. Conducting them is. My flow is nearly always the same: most things start with Fathom transcribing a meeting or Whisper Flow catching me talking out an idea, so I begin from a rich starting point. Then I use pull prompting: instead of telling the AI how to reach the destination, I tell it where I want to go and ask it to interview me, the way you hand Google Maps a destination and let it work out the route. I run that through Claude with my business brain plugged in, so every output already knows my voice, my clients and my work without me re-explaining.

That brain is the real unlock. Package your business knowledge into readable files any model can access, keep them on Google Drive or SharePoint, and you become tool-agnostic: when a better model lands you plug in and carry on, which is a big part of how you become an AI-first company. Master a handful of tools rather than hoarding, and run the same task through two to see which fits. If you want the fastest places to begin, start with the seven AI quick wins.

Final words

So don't try to own the whole toolbox. Sort your work into the six buckets, hold one tool per bucket, learn it properly, and hire the rest only when a real job calls for them. Keep your knowledge in a place you own so you can move between tools freely, and remember that none of this is here to replace your team. Used well, it takes the repetitive admin off good people so they spend their time on judgement, creativity and clients: the work that actually grows the business.

If you aren't sure where your business should start, book a free AI audit: we take your real workflows and hand you a prioritised next step. You can also watch the rest of AI Week.