Overview of The AI Tools I'm Building to Replace My Spreadsheets with Anish Acharya
In this episode, Chris Hutchins and Anish Acharya dig into how AI is replacing the spreadsheets, research workflows, and manual decision-making that power their personal and professional lives. The conversation centers on building practical AI tools for travel, health, finance, work, and family life—using Claude Code, OpenAI Codex, connectors, and custom workflows to automate repetitive tasks, surface better decisions, and save time. They also explore the bigger question: whether personal AI will make everyday consumers more optimized, more connected, or simply more overloaded.
Main Ideas and Takeaways
AI is replacing spreadsheet-driven decision-making
Chris describes how he used to plan trips by manually comparing Google Flights, award search tools, cash fares, and points redemptions in spreadsheets. Now, he’s building tools that do that work automatically:
- Search multiple travel sources
- Apply his personal rules
- Rank options based on points, cash value, timing, and convenience
- Return a concise recommendation in minutes
The broader theme: anything that used to require research, tabulation, and comparison can increasingly be delegated to an agent.
The real breakthrough is workflow design, not just chat
A major insight is that successful AI use depends on structuring the work correctly:
- Break tasks into discrete steps
- Use connectors for data sources
- Avoid trying to cram everything into one giant prompt or one endless thread
- Treat AI more like an operating system for repeatable jobs than a human teammate
Chris and Anish argue that the best results come from designing systems around how models actually work, not how humans work.
“Skills,” “connectors,” “projects,” and “channels” are the new building blocks
Chris lays out the architecture he’s using:
- Skills: repeatable tasks like flight search, transcription, invoice reconciliation
- Connectors: access to tools like 1Password, Google Drive, Apple Health, Gmail, etc.
- Projects: broader ongoing contexts like travel or finance
- Channels: ways to trigger AI through Slack, iMessage, email, apps, or dashboards
- Delivery: how the output gets returned
The key shift is away from a single agent with endless context and toward modular, reusable capabilities.
Personal AI is becoming a “zero marginal cost of work”
Anish frames this as a major societal shift:
- Software already gave us zero marginal cost of distribution
- AI now brings zero marginal cost of doing digital work
That means tasks like:
- monitoring prices,
- appealing bills,
- researching health options,
- managing cards,
- drafting communications,
- and finding obscure information
can increasingly be done for everyone, not just the highly organized or wealthy.
Practical Examples Discussed
Travel and points optimization
Chris’s travel workflows are a core example:
- Search award availability across tools
- Compare cash vs. points
- Apply personal preferences like:
- no very early flights unless the savings is worth it
- no layovers unless the discount justifies it
- prefer nonstop flights when available
He’s using this as a proving ground for how AI can replace hours of manual research.
Health data aggregation
The episode touches heavily on health as a future AI use case:
- Personal biomarkers
- Doctor records
- Apple Health data
- Private podcasts, journals, and research
- Trusted sources of expertise
The ideal system would combine:
- public knowledge,
- private/personal data,
- trusted expert sources,
- and AI that can synthesize it all.
Family and relationship communication
One of the most interesting examples is using AI to improve communication:
- Chris sends transcripts of conversations to the model
- It gives feedback on how to communicate better
- He also experimented with generating a family montage using photos and iMessages
The result is both practical and emotional: AI can help create richer, more meaningful artifacts from personal history.
Limitations and Cautions
Current AI is best at repetitive tasks
Both Chris and Anish are clear that today’s models still have limits:
- They make mistakes
- They need lots of instruction
- They struggle with ambiguous, high-stakes, or deeply contextual decisions
- They are not reliable enough to fully replace experts like doctors
Their rule of thumb: use AI to do the repetitive work, then sanity-check with humans where needed.
Context and memory are still messy
A recurring point is that models are only as good as the context they’re given:
- If the relevant data isn’t in the session, they don’t know it
- Long, mixed-purpose threads become confusing
- Separate tasks and clean workflows generally produce better output
Cost and subsidies will matter
Anish warns that today’s AI pricing is still heavily subsidized:
- $20, $100, and $200 monthly plans
- Plus API usage that can be far more expensive than users realize
Eventually, consumers will have to make real choices about:
- what tasks are worth paying for,
- which models to route to,
- and how much automation they actually want.
Bigger Philosophical Questions
Will AI make us more connected or more isolated?
The episode raises a thoughtful tension:
- AI can improve communication, planning, and shared experiences
- But it could also create “virtual intermediaries” between real people
Chris and Anish lean toward the optimistic view: for now, AI is mostly helping people communicate better and freeing time for real life.
Who benefits most?
A skeptical point comes up: maybe the same high-agency people who already optimize everything will benefit most from AI. The counterargument is that better interfaces, stronger prompts, and proactive assistants could help lower-agency users even more—by externalizing the work and nudging them toward better outcomes.
The system may have to change around us
Anish suggests AI will not just optimize individuals; it may eventually force institutions to adapt:
- banks
- insurers
- health systems
- travel providers
- government services
If everyone has an agent fighting bills, checking prices, and filing claims, the surrounding systems will need to become more responsive.
Actionable Advice
For people getting started
- Use Claude, ChatGPT, or Codex in their default apps first
- Start with repetitive tasks, not your most important decisions
- Break workflows into small steps
- Connect your tools gradually
- Use existing connectors before building custom infrastructure
For more advanced users
- Build with portability in mind: don’t tie your workflow to one model or one vendor
- Use a separate credential vault or service account for agent access
- Design tasks so they can be repeated, resumed, or handed off
- Keep human oversight for high-stakes areas like health and legal decisions
Bottom Line
This episode is less about abstract AI hype and more about the practical mechanics of replacing spreadsheets with intelligent workflows. Chris and Anish argue that the future of personal AI is already here in early form: a system that can search, compare, summarize, draft, and act across your life. The main challenge now is not capability—it’s architecture, trust, cost, and knowing which parts of life are ready to be automated.