Kevin demonstrates ways you can adopt AI in the next 100 days

Summary of the Podcast

Overview

• Episode 2 of a mini-series within The Next 100 Days Podcast where Kevin Appleby and Graham Arrowsmith discuss how AI is influencing their work
• Pre-recording conversation touched on football transfers (Nick Pope, James Trafford) before the formal recording began
• Kevin identified four or five distinct areas of AI usage he wanted to cover

Claude as a Personal AI Assistant (Pre-Call Preparation)

• Kevin's primary personal AI tool is Claude Cowork, which he connects to email, Notion, HubSpot, calendar, Dropbox, and Google Drive
• Use case: 10 minutes before a client call, ask Claude to summarise all recent interactions across those sources — it surfaces outstanding promises, discussion topics, and what to raise
• Claude can also draft emails into Gmail drafts (not sent), giving a ~90% ready version the user can refine
• Graham noted concern about integrating a Synology server; Kevin noted most personal data lives in Dropbox or Google Drive and Claude can connect to both

AI Agents Running Automatically

• Kevin runs two scheduled agents without manual triggering
• Agent 1 (daily): Scans calendar 30 days out and checks whether each diary entry has a corresponding task in Notion; creates missing tasks automatically, including a Notion page for notes
• Agent 2 (weekly): Scans upcoming 30 days for GrowCFO Show or Next 100 Days podcast recordings, researches the guest online, and appends notes (bio, website links, suggested questions) to the relevant Notion task
• Kevin deliberately ignores AI-suggested questions to preserve natural curiosity in conversation, but acknowledges suggested questions are useful for solo-host formats
• The same agent concept applies to client/partner meetings: an agent could automatically pull last interaction notes before any key meeting

Adopt AI in Finance Operations (GrowCFO Context)

• GrowCFO's current quarterly theme is Intelligent Finance Operations — covering the end-to-end finance cycle from purchase orders to final accounts
• Three-way matching: AI can match purchase orders, goods receipt notes, and invoices automatically; if all match, the invoice is paid without human intervention
• Fraud detection: AI spots anomalies far more effectively than manual review
• Reporting: AI can generate board reports and dashboards from accounting systems; demonstrated in a GrowCFO webinar with tech partner Round Treasury using Claude
• Key value: AI removes grunt work so finance professionals can focus on what the numbers mean, not just crunching them

AI's Impact on Finance Jobs

• Significant job displacement expected in transactional/lower-level finance roles; less so for senior finance business partners whose work is relationship-based
• Challenge: AI often saves portions of multiple people's jobs rather than eliminating whole roles, making headcount reduction difficult — a pattern observed in shared services projects long before AI
• Supply of finance professionals is falling, particularly FP&A specialists in the US; AI may initially ease recruitment pressure rather than cut headcount
• Most finance teams are still at the "chatting with ChatGPT" stage, not yet using agentic or Cowork-style tools
• A live example of a fully agentic debt-chasing system: an AI agent that checks the debtors ledger, identifies overdue accounts, calls customers, and holds an intelligent conversation about outstanding invoices

AI Connectivity Limitations

• Claude's current integration with Xero is poor — only capable of producing basic debtor reports rather than actionable overdue lists
• Copilot and Gemini do not support Dropbox connectivity, limiting their use in Kevin's personal setup
• Hallucination risk is reduced when AI operates within a well-defined, high-quality data "cocoon" — the importance of a single source of truth

Faster Close and Rolling Forecasts

• "Faster close" — producing monthly accounts quickly rather than 10–15 days after month-end — has been a finance aspiration for 20+ years and AI now makes it achievable
• Rolling forecasts with multiple scenarios (e.g., 0%, 25%, 50% tariff scenarios) can be modelled and run instantly by AI rather than taking weeks to build

AI Governance in Finance

• Segregation of duties must be replicated in AI workflows: different agents or human approvers for setting up suppliers, authorising accounts, and making payments
• A single "super AI agent" handling everything end-to-end is not yet appropriate or auditable

Using AI for Research & White Papers

• Kevin used ChatGPT Deep Research to read ~30 published documents from major consultancies (PwC, Deloitte, EY, Accenture, BCG, Forrester, etc.) on AI in finance ops in 15–20 minutes — work that previously would have taken two graduate trainees a fortnight
• He gave both ChatGPT and Claude the same prompt and found ChatGPT produced a better overall result, though Claude surfaced some content ChatGPT missed; he combined both
• AI is excellent at generating and researching text but structuring the final document, choosing diagrams, and making it consumable remains a human task

Virtual Board / AI Avatars

• A GrowCFO partner CFO used AI avatars of their board members to anticipate board reactions to finance reports before presenting
• Kevin tested a "virtual advisory board" in AI, including public figures like Michael Heppell, and received useful diverse perspectives
• The deeper goal: frame board materials to open up discussion rather than trigger defensive reactions

KAIOS — Kevin's AI Operating System (Personal Project)

• Started from a ChatGPT conversation: "What if everything you've been taught about time management is a lie?" — leading to the insight that most productivity systems just create more tasks rather than freedom
• Evolved into KAIOS (Kevin's AI Operating System): a Dropbox-based knowledge structure containing Kevin's values, StrengthsFinder profile, career history, stories, and frameworks — accessible by any AI model via instructions stored within the structure
• A book outline, introduction, and chapter frameworks have been drafted with ChatGPT's help; the project paused in favour of building KAIOS first so the book has genuine personal stories and IP embedded
• Graham suggested this could be Kevin's biggest career opportunity — analogous to how MeclabsAI has built a multi-million pound business around systematised marketing knowledge
• Kevin noted he prefers positioning himself as an expert using AI in finance rather than as an AI expert per se, given how fast the field moves

Career Reflection & Future Direction

• Both hosts reflected on the retirement vs. continued work question; Kevin expressed enthusiasm for staying engaged with AI developments and not wanting to reach the point where technology no longer makes sense
• Kevin acknowledged competitors in the AI-for-finance space (e.g., Nicholas Boucher's AI Finance Club) but sees his differentiation in applying deep finance and personal IP rather than competing directly
• GrowCFO Show is approaching episode

The Next 100 Days Podcast Co-Hosts

Graham Arrowsmith

Graham founded Finely Fettled in 2014 to provide data from The UK High Net Worth Database to marketers targeting affluent and high-net-worth customers. He's the founder of MicroYES, a Partner for MeclabsAI, creating lead generation AI Agents & Workflows and introducing the MeclabsAI Platform. Graham is an inCruises Independent Partner, and is building up interest from people around the world in the World's Largest Travel Membership - inCruises. You can sign up and access 21,000+ cruises, hotels and tours by clicking HERE

Kevin Appleby

Kevin specialises in finance transformation and implementing business change. He's the COO of GrowCFO, which provides both community and CPD-accredited training designed to grow the next generation of finance leaders. You can find Kevin on LinkedIn and at kevinappleby.com