Better Execute · Essay

Why Strategy and Execution Are About To Diverge — And How LLMs With Full Company Context Become The Unlock

By Michael Urness — Founder, Better Execute

A real-world example of how leadership teams using full-context LLMs make dramatically better strategic decisions — and why the rest of the market is about to fall behind.

After 25 years working on why some companies execute well and most don't — as operator, investor, advisor, and now founder of an AI-native execution platform — I am convinced full-context LLMs are the biggest single decision-making unlock I have ever seen. Roughly 10x better than anything before. Not a feature. A way of working — and one of the easiest to miss because it doesn't look like a feature.

Let me show you what I mean with one real afternoon of work.

A short, real example

Recently, a real strategic question landed on my desk:

Our customers love DCE for strategy and execution, but they keep one foot in Asana for day-to-day task management. Should we build Asana-style task management into DCE? If so, how — and how do we avoid becoming "Asana with a strategy tab"?

Two paths existed:

1. The traditional path. Brief a product manager or consultant. Wait a week or two. Receive a feature comparison spreadsheet, a phased roadmap, and a build estimate. Begin building.
2. The new path. Ask an LLM that had been given the company's macro strategy and the company's micro state — its codebase, its data model, its agent architecture, its customer constraints, its non-negotiable product rules — to produce the strategy.

I took the second path. By end of day there was a complete planning artifact in the repository — a feature comparison, a gap analysis, a strategic synthesis, a phased plan, a 4-week sprint, risks, open questions, and explicit guardrails — and the analysis was better than what a senior product hire would have produced in a week, because it could synthesize across layers no human can hold simultaneously.

That same artifact is bundled with this essay as a downloadable companion document. The LLM saw what I saw, plus everything I didn't have time to re-read, and produced a synthesis I genuinely could not have produced alone in a day.

This is the story I want to tell you, because I believe it is the biggest unlock for company decision-making and execution we are going to see this decade — and most companies will not figure out how to operate this way.

The naive answer, and why it would have been wrong

If you asked an outside consultant — or honestly, most product people — "should DCE add Asana-style task management?" you would get a roadmap that looks something like this:

Asana has subtasks, custom fields, multiple assignees, attachments, comments, Kanban, calendar, Gantt, portfolios, forms, rules engine, dashboards, and AI. Build the gap. Here's a 9-month plan.

This answer is rational. It is also wrong — and it would have steered the company straight into the trap. Building toward feature-parity with Asana would have made DCE generic, which is exactly the failure mode our company strategy is built to avoid. We would have spent six months and arrived at "a worse Asana."

Spotting that requires holding two things in mind at the same time:

• The macro — the company's actual strategic position, the reason it exists, the things it must never become.
• The micro — the actual schema, the actual code, the actual customer constraints, the actual product stage.

Most decision-making fails at the seam between those two. People with macro context (founders, board) don't know the micro. People with micro context (engineers) can't fluently apply macro logic to micro decisions. The LLM is the first thing in business history that can hold both layers at once and reason between them — and that single capability is what changes the economics of strategic decision-making.

What the LLM was given

When I asked the question, the model could already see:

Macro context. Our company's product principles document — written in plain English — including:

• Do not make DCE generic.
• Do not silently change terminology in the UI.
• Customer-specific requests should be generalized carefully before becoming core product features.
• The strategic idea is that humans and AI agents share visibility, accountability, and work inside the same execution operating system.
• Per-org configuration is the default isolation mechanism. Customer-specific branches are not acceptable.

Micro context. The actual codebase: the data model, the agent architecture (with its proposal/evidence/confidence pattern), the existing notification system, the existing strategy plan with its 13 ordered sections, the existing dependency model that already spanned plans/milestones/todos but wasn't enforced.

Constraint context. Where the product was in its lifecycle, which customers were active, and what kinds of changes carried the highest risk right now.

Policy context. That every cron route must fail-closed when its secret is unset. That every PR must include a release note. That any change touching permissions, data, or terminology requires explicit risk flagging.

No outside consultant could absorb that volume in a week. Most product hires take a quarter to internalize it. The LLM had it on tap, in seconds, every time it reasoned about a tradeoff.

What came out

What came out of that single half-day session was not a feature list. It was a synthesis.

The strategic synthesis

Asana model: humans manage work; AI is a sidebar that suggests when invited.

DCE model: humans and agents co-manage work; the agent canvas is a peer of the human canvas, not a feature of it.

That single framing is the entire competitive thesis. It came from the LLM cross-referencing the macro principle "don't make DCE generic" against the existing agent-proposal architecture in the codebase. A consultant looking only at feature gaps would have missed it entirely.

From there, five differentiators were named — each rooted in a specific capability the codebase already had but hadn't yet expressed as product:

1. A continuously-curated personal inbox that routes work to attention with strategic causality.
2. A strategy-causal task graph — every todo can answer "what Rock does this serve, what priority, what KPI?" and agents can traverse and infer those links.
3. Proposal-driven mutation as the default write path — every agent change has trigger, evidence, confidence, and a human approval gate.
4. Meeting-native work creation — transcripts become decisions become todos with owners and dependencies, all governed.
5. Predictive Rock health — agent-driven trend analysis that flags at-risk priorities before the quarter ends.

These are not Asana features. Asana cannot easily build them, because they assume an agent-native data model and a human-approval governance loop that Asana's product was never designed around. The wedge is not a feature. It is the governance model.

That is a defensible competitive position. It came out of the model in one pass.

The recommendation

Ship the table-stakes floor so leadership teams can genuinely live in DCE without an Asana tab open. Skip Asana's depth ceiling — timeline, rules, proofing — until customer demand is concrete. Pour the differentiation budget into the agent-native wedge.

Crisp. Decisive. Aligned with the macro principle. Aware of the build-vs-defer tradeoff. Disciplined about not chasing scope for its own sake.

The 4-week plan

Then it produced a concrete, week-by-week sprint plan that:

• Re-used the existing Notification model rather than introducing a parallel one (because it had read the schema and knew the model already existed).
• Pre-baked an authorType field on the new Comment model so AI agents could later post comments through the same primitive — a five-minute schema decision today that saves a six-week migration later.
• Renamed the page "My Work" not "My Tasks" — because terminology preservation was a macro-level non-negotiable.
• Explicitly sequenced itself around the current product stage — because timing was a macro-level constraint that would re-order the picks.
• Caught its own oversight on the Notification model when an automated reviewer challenged it, and corrected the doc the same hour.

Every micro decision in that plan was traceable to a macro principle. Every macro principle was applied with knowledge of the actual code. That is the synthesis layer that almost never happens cleanly inside a normal company.

Why this matters for your business

The work above is not extraordinary. It is going to become normal — for the companies that figure this out.

What it requires is two things.

First, a company has to actually write down its macro strategy in a form an LLM can read. Not a deck. Not a lunchtime pep talk. Not a private spreadsheet. A versioned, structured document the model loads as context every time. We keep ours in a single canonical document at the root of the codebase, but the format does not matter — what matters is that the company's strategy, its non-negotiables, its terminology, its current stage, and its operating rules are legible to the model.

This is harder than it sounds. Most companies' strategy is implicit — it lives in a founder's head, in three slide decks, in a private Notion. To get the unlock, the strategy has to come out of those places and into one canonical document. The act of writing it down is itself clarifying. Many of the leaders I've worked with over the years have told me afterward that the writing-it-down step alone was worth more than the AI tooling that followed.

Second, the LLM has to be given micro access — to the data model, the source code, the customer feedback, the actual operating reality. Not a summary. The thing itself.

When both of those conditions are true, the LLM becomes the first practical mechanism in business history for fast, repeatable synthesis between strategy and execution. Not "AI summarizes a meeting." Not "AI drafts an email." Real strategic synthesis — the kind a great chief of staff produces — but available on every decision, every day, in minutes.

What that means competitively

A traditional company's decision-making latency on a question like "should we add Asana-style features?" is measured in weeks. You convene a meeting. You assign a PM. You commission a competitor analysis. You wait for a draft. You workshop it. You revise. You decide. By the time the decision is made, the market has moved.

A company operating with full-context LLMs will resolve the same question in hours, with a higher-quality answer, and ship a working sprint plan the same week. Then it will re-resolve it as conditions change — because re-running the analysis is also cheap.

This compresses decision cycles by 10–50x at the strategic-synthesis layer.

But the more important effect is quality. A company without this capability decides on whichever framing happened to be in the room. A company with this capability decides with every relevant constraint considered, every macro principle applied, every micro detail accounted for. Bad strategy gets caught before it becomes bad code.

Companies that figure this out will operate as if they have a senior product strategist, a senior architect, a senior PM, and a senior compliance reviewer in the room for every decision — because they will. Not as a metaphor. As an operating reality.

A traditional company cannot keep pace with that. Not because it lacks talent — but because the synthesis layer is a physics constraint of the way humans work. One person cannot simultaneously hold the company's macro principles, the codebase's micro state, the customer constraints, and the regulatory rails in active working memory while also reasoning about a tradeoff. Two people can do it together over a week. An LLM with full context does it in 90 seconds.

The companies that move first will not just be faster. They will make systematically better calls, because the cost of checking a decision against the full context drops to roughly zero. Every decision gets pressure-tested. Every plan ladders to the macro. Every micro choice is interrogated against company principles. Drift gets caught before it ships.

The companies that don't — most of them — will continue making their decisions the way they always have: in rooms, on intuition, with whichever subset of context happened to be present. They will lose, slowly at first, then quickly.

What the case study really proves

I started this doc with a small, real example: a single-day decision about whether DCE should chase Asana on task management.

What that session actually demonstrated is not "AI is good at writing docs." It demonstrated that:

• Strategic clarity went up, because the LLM articulated the wedge ("governance model, not feature parity") before any human had named it.
• Execution detail went up, because the same LLM produced a 4-week sprint plan grounded in the actual schema and real product constraints.
• Macro–micro alignment went up, because every micro decision in the plan was traceable to a written macro principle.
• Self-correction worked, because when an automated reviewer challenged a model assumption, the doc was corrected in the same hour with no ego in the way.

That is the new operating loop. Macro context in. Micro context in. Decision out. Reviewer challenges. Correction in. Plan committed. Execution begins. Repeat.

It is the closest thing I have seen — across 25 years of working on this problem — to a genuine compounding advantage in how a company operates. Most "productivity" gains in that quarter-century have been incremental. This one is not.

The actual planning artifact — gap analysis, wedge synthesis, phased LOE, 4-week sprint, risks, open questions — is downloadable alongside this essay. It is not polished marketing copy. It is the working planning document a senior team would have spent two weeks producing. It was produced in an afternoon.

That is the unlock. Whichever companies adopt this way of operating will pull away from the rest faster than the rest will realize what is happening.

If you want to talk through what this would look like in your company

If you're trying to figure out where this could change how your team makes decisions — what your version of the macro-context document looks like, where the model would actually plug in, what's worth tackling first — I'm happy to spend 30 minutes thinking through it with you. No prep required, no slides, no commitment after. Just a conversation about what you're trying to do and whether this approach fits.

Book a 30-minute conversation →

I've spent 25 years watching companies try to execute well. This is the one shift I would not want a leader I cared about to miss.