My 3 interesting things for you this month…
1. “Something big is happening” and what it means for finance
Matt Shumer’s article has already had 80M+ views, so you may well have read it, and it’s doing the rounds for good reason.
His core point is simple: the gap between public perception of AI and what the best models can do right now has become enormous and that gap is going to catch a lot of people out.
Two parts of his argument are particularly relevant for finance:
1) Most people are judging AI based on the free tier
Shumer makes the point that ‘free’ versions of AI are already running a year behind paid tiers. He argues that if you’re only using whatever is free, you’re not seeing the advanced capability that paid users are already building workflows around.
2) The pace of improvement is now the headline
He also outlines just how fast AI is improving. Capability is stepping up significantly every couple of months, to the point where last year’s experience of using AI already feels outdated.
This is especially true for screen-based knowledge work, which makes up a huge proportion of what finance does: analysis, modelling, commentary, decks, reconciliations, drafting papers, writing narratives, synthesising information for stakeholders.
So if capability keeps compounding, finance is absolutely in the blast radius.
The other piece that has spooked markets recently is Citrini Research’s The 2028 Global Intelligence Crisis.
It’s written as a fictional backward-looking memo from 2028, mapping out a scenario where “abundant intelligence” drives a sharp economic adjustment, rising unemployment and a meaningful market drawdown. Citrini is explicit that it is a scenario, not a prediction.
I mention both pieces together because they sit on the same spectrum:
- Shumer is saying: “The capability is here, most people are behind and you should act.”
- Citrini is saying: “If that trajectory plays out fast, how would you prepare for this scenario.”
My take as it relates to finance leaders
I don’t think anyone can say exactly how fast this will hit each sector, and I’m a bit more hesitant to make big sweeping predictions on timing, especially about finance.
Even the market debate right now is partially about speed: how quickly AI moves from early adopters to broad daily use at work.
That said, the direction of travel Shumer describes is very close to what I’m seeing with finance leaders who are genuinely experimenting: real pieces of work getting done faster, with fewer iterations, by people who have learned how to brief and check these tools properly.
Where I’m very aligned with him is on the action he recommends. The leaders who do best from here won’t be the ones reading the most think-pieces. They will be the ones who are:
- Using the strongest models available, not just the free tier
- Pushing AI into real work (process diagnostics, speeding manual work, improving outputs), not just asking it trivia
- Building the habit of experimenting and learning in public with their teams
That is the pattern I’ve been helping FDs and CFOs put in place: use AI to raise the bar on your own work, fix broken processes and create capacity for the parts of finance that still absolutely need human judgement.
If you’re a finance leader, Matt’s article is worth a quiet 10–15 minutes with a coffee. You can read his piece at his site and the Citrini piece, and decide for yourself.
And if you’d like to talk about what “getting ahead of this” looks like in a finance function, I’m always happy to chat.
2. Beyond the Beta: Claude is now in Excel and PowerPoint
Let’s take a look at something more concrete and already happening.
For most finance teams, AI has been living in an extra tab. You jump out of Excel, ask a question, paste something back in, then spend the next ten minutes checking formatting and wondering if it made up the answer.
Anthropic is trying to remove that friction by putting Claude directly inside the tools people already use, including Excel, PowerPoint and Slack via a new package including Cowork and Plugins for Enterprise.
It’s part of the broader “AI is moving into the workflow” trend, where the value is not just a clever chat response, but real analysis and spreadsheet or slides work that improves speed, accuracy and decision support.
Two things from this:
1) The AI race is shifting from who has the best AI models to who has the most useful distribution
Anthropic is positioning Claude as a “workplace layer”, competing with Microsoft Copilot, Open AI and Google’s Workspace efforts. The headline feature is less “here’s a smarter chatbot” and more “here’s an assistant that can works where the work is.”
They’re leaning into connectors (e.g. Google Drive, Gmail and DocuSign) and department-focused plugins, including finance use cases.
Note that Claude can’t currently plug directly into M365. For now, it can work in individual files, but not across all your files and email at the same time – yet.
2) The pace has picked up and it is starting to show up in day-to-day tools
This is the bit you can feel. Capability isn’t just improving – over the last 2 months it looks to be an order of magnitude more useful – but it’s also getting embedded in what we actually use.
When AI sits directly in Excel, the gap between “experimenting” and “changing how we work” gets much smaller. A easy way to think about this is it’s doing what you wish Copilot would do.
Two things to keep in mind
That all said, there are two big caveats worth calling out explicitly, especially for finance.
Caveat #1: Adoption will be uneven and slower than the headlines suggest
Some enterprises will move carefully because of security, procurement, data access and audit expectations. Some SMEs will move slowly because they don’t have the time, budget or internal champions to rework processes.
The result is a messy middle where AI is available, but not consistently used and not consistently governed.
Caveat #2: You cannot delegate the job, you have to partner with the tool
If Claude can draft an analysis inside Excel, it can also confidently produce something that looks right but is built on the wrong assumptions. Finance still owns judgement: the inputs, the logic, the reconciliation and the story.
The winners will be the people who go beyond “using AI” and can direct it well, verify it thoroughly and explain the output clearly.
Putting this into action
A practical takeaway to share with your team this week: pick one recurring task (variance commentary, management pack charts, first-pass forecasting notes) and define the partnership:
- What does the AI do first?
- What must a human always check?
- What evidence do we keep so the output is defensible?
If AI is moving into Excel, “AI skills” stops being a nice-to-have and becomes part of core finance competence.
If you’d like to see a demo of Claude in Excel and help your team think through how they could start integrating these skills before your organisation buys the tools, get in touch about my virtual 3-hour Making AI a Daily Habit team course.
3. Claude Skills are the quiet feature that makes AI useable at scale
And finally, as many of you are new to Claude, here’s something to try and have a play with.
Most of us have had the same experience with AI so far.
It can be brilliant… and then wildly inconsistent. Two people ask the same question in slightly different ways and get different outputs. Or you get a great answer once and then spend the next month trying to recreate it.
Anthropic’s answer to that problem is Claude Skills. In simple terms, a Skill is a packaged, reusable workflow that teaches Claude how you want a task done, so you don’t have to re-explain the process every time.
So what are Skills, exactly?
Anthropic describes Skills as folders of instructions, scripts and resources that Claude can load to improve performance on specialised tasks.
There are three flavours worth knowing about:
- Anthropic Skills(built-in) like enhanced document creation for Excel, Word, PowerPoint and PDF. These can be invoked automatically when relevant. Last week, in just 70 minutes, I used these Skills to create an Excel model, a PowerPoint that explained it and an accompanying Word doc.
- NOTE: these skills are still not reliable enough for your finance teams ongoing needs, but you can start seriously experimenting with them right now.
- Custom Skills allow you to create for your own repeatable workflows (personal, team or company-specific).
- Organisation skills (Team and Enterprise) where owners can distribute approved Skills across the business.
How do they work?
The clever part is that Skills are designed to avoid the classic “everything in the prompt all the time” problem.
Anthropic calls this progressive disclosure: Claude only loads what it needs when it needs it, rather than dumping a whole manual into the conversation context. This results in consistency without stuffing every chat full of process detail.
Why finance should care
If you zoom out, this is the shift from “AI as a chatbot” to “AI as a workflow partner”.
The Claude Skills approach show finance examples like building financial dashboards in Excel, generating an executive PowerPoint from the numbers and automating multi-format reporting pipelines (Excel + PPT + PDF).
This matters for three reasons:
1. Repeatability becomes a thing
Instead of hoping someone remembers the right prompts, you can codify the workflow once, then reuse it.
2. Standards are easier to enforce
Skills can embed best practice, structure and quality checks in the workflow itself, not just in someone’s head.
3. This is a genuine career differentiator
A lot of the AI conversation is still vague. “I use AI” is not a skill. Being able to say “I designed an AI-enabled month-end workflow that improved consistency and reduced cycle time” is much closer to real value.
It links nicely to the career lens I bang on about a lot: what problems are you solving for others?
Three caveats (because it’s finance)
1. The tools are very new. People are still finding the edges in them. It’s too early to either rely on them, or promise ROI if the team adopts them. That said, I expect ROI by the end of this year.
2. This doesn’t remove accountability. It makes it easier to produce work, but you still own the judgement, the checks and the narrative. This will be the key differentiating skill of the future.
3. Adoption will be uneven. Team and Enterprise tooling helps standardise, but plenty of SMEs and cautious organisations will move slower for governance, security and change reasons. Finding safe ways to practice and build skills are key.
One practical action for this week
Pick one workflow that is repetitive and annoying (board pack charts, weekly KPI email, month-end narrative).
Write down:
- The inputs you want used
- The checks that need to happen before anything goes to a stakeholder
- Ask an AI tool (even Copilot or ChatGPT!) how a Claude skill could help you with this workflow, and what you would need to do to get it to work
Even if you never build a formal Skill using this, you have just created a mini operating procedure for “AI + human” work. And if you do build one, you are setting yourself up for consistent output across the team.
Want more insights?
Sign up to my interesting things newsletter to get tips like this and more.
