Get ROI from Your AI: Quantifying AI Value

The tools I am actually using, five agents worth knowing, and the uncomfortable truth about why most people are not getting results.

I will be honest with you. For the first six months I had access to AI tools, I used them almost exclusively to write emails I was too lazy to draft myself. Not bad emails, actually pretty decent ones. But that is roughly the equivalent of buying a Ferrari and using it to collect the kids from school.

Most of the people I speak to are doing the same thing. They use AI as a slightly smarter autocomplete, get marginal time savings on tasks that should not take long anyway, and then wonder why the return feels underwhelming. According to Google Cloud's 2026 Business Trends Report, employees using AI agents are saving an average of forty minutes per interaction. Per interaction, not per day. The maths gets serious very quickly.

Tip 1: the five minute competitive intelligence brief

In the corporate world you have analysts. In a startup you have yourself and a guilty feeling that you should know more about your competitors than you do. Perplexity does in five minutes what used to take an afternoon. It pulls live web data and cites its sources, so every claim links back to where it found it. When you need a quick read on what a competitor has been saying publicly, funding announcements, hiring patterns, new product pages, recent press, it delivers.

A sales director I know runs a lean team at a clean energy company. Before every enterprise pitch she runs a ten minute brief covering three things: the prospect's stated sustainability commitments, their recent hires in the relevant function, and what their CEO has said about the category in the last ninety days. Her second meeting conversion is sixty eight per cent.

Tip 2: turn your meeting notes into CRM updates in under two minutes

This sounds boring until you add up the hours. Fifteen to twenty minutes after every significant meeting writing notes, updating the CRM, drafting a follow up. Call it eighteen minutes, five meetings a week, four people. That is six hours a week of admin that creates no revenue.

The fix is almost embarrassingly simple: voice note plus AI. Record a two to three minute voice note, no structure needed. A transcription tool turns it into a tidy output, then a single prompt structures it into a CRM update and a follow up draft. The real value is not just time. It is capture quality. When you write notes manually you filter as you go, keeping the big things and losing the small signals. Three months in, I found a cross sell opportunity in a voice note that I would never have written down. That single deal was worth more than a year of every tool I use.

Tip 3: build a knowledge base that actually knows your business

Generic AI is smart. AI fed your documents, your past proposals, your customer conversations and your market analysis is a different animal. Google's NotebookLM does this better than anything else I have tested, and the core version is free. You upload your own material, then ask questions against it. It answers based only on your documents, with citations down to the paragraph.

When I write a proposal for a prospect I have worked with before, I ask it to pull relevant points from previous conversations and related case studies. It once surfaced a statistic I had written eight months earlier and completely forgotten, and that stat became the centrepiece of the proposal. It also gets better over time. Every document you add makes it richer. After three months mine knew more about my business context than most of my colleagues.

Tip 4: the weekly thirty minute time audit

A founder I respect, in sustainable packaging, started doing a weekly AI audit. Every Friday she exports her calendar into a chat tool and asks it to sort everything into four buckets: revenue generating, relationship, operational necessity, and distraction. The first time she ran it, over forty per cent of her week was in the last two. Over six months she moved her revenue generating time from around twenty per cent to just over thirty five. The AI also flagged a pattern: three weekly internal alignment meetings could become a single async update, saving her team over two hours a week.

If you only try one thing from this entire article, try this one. It is the least exciting and the most impactful.

Tip 5: use Claude to build MVPs, apps and websites without writing code

Claude can build a working website, a functional app prototype or a complete landing page from a single conversation. Not a wireframe, a working, deployable thing you can send to a client, test with users, or launch on your domain. I have used this for internal tools, landing pages for new offerings, calculators for prospects, even the content planning app I use to manage these newsletters.

Last month I needed a landing page for a new Modus Verde service. Normally I would brief a designer, wait three days, go through two rounds of revisions and pay between five hundred and fifteen hundred pounds. Instead I opened Claude and twenty minutes later had a polished page. It has been live for three weeks and captured forty seven email signups. The quality of the output is directly proportional to the specificity of your brief. Spend five minutes writing a detailed one and you save five hours on the other end.

The uncomfortable truth

Here it is. The reason most people are not getting real return from AI is not the tools, the prompts, or the learning curve. It is that getting proper value requires you to be specific about what you actually want. And being specific requires you to have thought clearly about what you are trying to achieve. That is the hard part of most work, and it was hard long before AI arrived.

The people getting serious return are not the most technically sophisticated. They are the clearest about their own objectives. They know what a good meeting output looks like, what a good proposal contains, and what a good week looks like for them. The AI is not smarter than your strategy. It is a fast, tireless executor of whatever direction you give it. Vague in, vague out. Sharp in, sharp out. Start there.

Tools mentioned

• Perplexity Pro, real time research with cited sources, around twenty dollars a month.
• Otter.ai, meeting transcription and AI notes, free or around seventeen dollars a month.
• Google NotebookLM, a knowledge base built from your own documents, free or with AI Premium.
• ChatGPT Plus, a versatile general assistant for research and triage, around twenty dollars a month.
• Claude Pro, deep thinking, MVPs, apps and meeting prep, around eighteen pounds a month.
• Gemini in Google Workspace, AI inside Gmail, Docs and Sheets, included in Workspace plans.