Entering the Markdowns Era

Why the future of working with AI is not better prompts, but better structure.

Why prompts feel flat

Is it just us, or does the era of prompts feel flat?

Not wrong. Not disappointing. Just flat. And we can't help but wonder what's coming next.

In 2024 and 2025, we collectively learned how to prompt. We experimented with frameworks that front-load context, instruct AI to ask clarifying questions, play a particular role ("you're a finance expert with 20 years experience"), and shape the output we want. For many people, this created a genuine productivity unlock, often estimated at a 10 to 30 percent improvement depending on role and task.

And yet, AI can still feel like an extra limb. Powerful, yes, but not fully integrated into how teams actually work. It makes us faster, but it is not always transformational. It doesn't bring all the moving pieces together in a way we know it should.

That raises a more interesting question. Not how do we prompt better, but how do we work with AI in a way that is natural, consistent, and scalable across teams?

Our belief is that the next step is not more prompting. It is a better way of working. It centers on using markdowns.

My stance on prompts

To be clear, this is not an argument against prompts. Prompts are useful, impressive, and often the first way people discover what AI can do.

But prompts have structural limits when we try to scale beyond individual use or across multiple sessions.

The first limitation is inconsistency. AI answers are derived from the context provided, and context depends on what someone remembers to include in that moment. Important nuance can be missing, outdated, or framed differently each time.

Second, variability. No two people prompt the same way. Similar questions can produce different outputs because assumptions and language differ.

While this is manageable for personal productivity, it's risky for shared decision-making. In short, prompts are episodic. Real work is cumulative.

From prompting to persistent context

This is where markdown files enter the picture.

I was recently introduced to the idea of using markdown files as persistent sources of truth — structured documents that capture the latest thinking on a topic and can be attached to AI conversations to drive analysis, insight, and execution.

The closest real-world analogy is a company policy, like a Delegation of Authority or Procurement Policy. But unlike traditional policies, markdowns aren't static documents designed to sit in a folder. They're living artifacts, written in clear language, updated as understanding evolves, and actively used in day-to-day work with AI.

What's important here is that markdowns don't replace prompts. They change what prompts do. Instead of spending your time recreating context, prompts become instructions layered on top of shared understanding.

Using AI to create the markdowns themselves

This is the part that often gets overlooked: you don't need to start with perfect markdowns. In fact, AI is extremely good at helping you create them.

You can use prompts to turn messy thinking into clean structure. You can ask AI to take meeting notes, policies, strategy decks, or even rough bullet points and synthesize them into a clear, well-structured markdown file. You can then review, refine, and agree on it as a team.

In other words, prompts are how markdowns are born.

This creates a natural progression. Early on, you prompt heavily to extract clarity. Over time, the outputs of those prompts harden into markdowns — reusable, shareable context that no longer needs to be reinvented every time.

Prompts are the scaffolding. Markdowns are the structure that remains.

Example: Large corporate transformation initiative

To put this theory into practice, let's consider an enterprise initiative with many people working together on different workstreams. Instead of relying on scattered prompts, decks, and emails, a transformation team can maintain a small set of markdowns. For example:

• About_Initiative.md - Sponsor goals, values, milestones, and key dates.
• Initiative_Governance.md - Stakeholders, RACI, governance structure, and constraints.
• Initiative_Budget.md - Budget position, performance, and forecast assumptions.

Now imagine the program receives a $500K change order driven by unresolved dependencies causing schedule changes. The project manager is now tasked with reviewing the change request, explaining the change tailored to her different audiences, scheduling the approval meetings, obtaining approvals, preparing communications, all while keeping the project moving forward.

Yes, this could be handled through a series of prompts. But that approach depends on one person accurately consolidating all relevant context into a coherent input — and having the time and cognitive capacity to do it well. The risk of omission is high. More importantly, it constrains the AI to the framing provided, rather than allowing it to surface its own insights, connections, and conclusions across the full body of information.

With markdowns, the AI already understands the program's objectives, values and key dates. It knows the goals, the people, and the financial constraints. The prompt becomes a trigger for analysis, not a test of memory and autonomous execution.

With these markdowns attached, the AI has a consistent and comprehensive understanding of the entire program. It can connect the dots between different audiences, upcoming meeting dates, budget constraints, program values, and strategic objectives — without anyone having to re-explain the context each time.

What this unlocks versus standard prompting:

Example: Startup launch

The same logic works for startups with a different markdown set. The context shifts from enterprise governance to speed, product clarity, and execution risk, but the principle is the same: keep critical context in a small set of living markdowns so AI can reason consistently across decisions. For example:

• Product_Aspirations.md - Product vision, growth targets, and market direction.
• People_&_Legal.md - Team roles, investors, and legal constraints.
• Technology_Stack.md - Systems architecture, integrations, and technical boundaries.

When the founder asks AI to build a product roadmap, the difference isn't in how clever the prompt is. It's in the quality and persistence of the context that surrounds it. AI understands the aspirations, growth, technology stack, people, legalities, and more, and can combine these into a cohesive strategy, prepare the plan, the team onboarding plan, and more.

This is how small teams move faster without losing coherence.

Early guidelines on an emerging practice

This way of working is still emerging, but some patterns are already clear.

Less is More

It's better to have fewer, higher-quality markdowns than many overlapping ones. Each markdown should have a clearly defined scope and act as a single source of truth for its topic. This is similar to team design: overlapping roles create confusion and gaps create risk.

Concise Markdowns

Markdowns should be concise without being shallow. Long documents aren't inherently bad, but unnecessary length creates bloat. Governance matters here - someone needs to own clarity and keep things sharp.

Leave Room for Iteration

Importantly, markdowns should evolve. As understanding deepens, they should be updated, refined, and occasionally restructured. This isn't busywork; it's how institutional knowledge compounds.

Natural language meets computer speak

Markdown works because it sits between natural language and machine-readable structure. Headings, bullets, and plain text tables create clarity for humans and reduce translation for AI.

Compared with slides, screenshots, and PDFs, markdown lets AI spend less effort interpreting format and more effort on reasoning and synthesis.

Less translation. More thinking.

# Transformation Program Overview

## Purpose
- Reduce operating cost while improving delivery speed
- Establish repeatable decision-making frameworks

### Decision rule
- Prioritize **customer value** over `local optimization`.

## Principles
1. Decisions favor long-term value over short-term optics
2. Speed matters, but clarity matters more

## Key Dates
- Program kickoff: March 1, 2026
- Steering Committee: Monthly
- Target completion: December 2026

Embedding AI into the core analytics engine

The real power of this approach is organizational coherence.

When teams share markdowns, AI becomes a consistent analytical partner rather than a personal productivity hack. Different team members can independently explore scenarios, test assumptions, and generate insights — all grounded in the same underlying context.

Compared to traditional prompting, this feels less like asking for help and more like collaborating with a well-briefed colleague.

That's how small teams start to operate like much larger ones with near infinite analytical power. And that's when AI stops feeling like an extra limb - and starts feeling like part of the team.

Conclusion

There's still a lot to learn.

Markdowns aren't a silver bullet, and they require discipline to maintain. But they represent a meaningful shift - away from episodic prompting and toward shared, persistent understanding.

Prompts helped us discover what AI can do. Markdowns help AI understand how we work.

And that feels like a promising place to build from.