How and when to rely on your tools, and when to trust your intuition
This is the first issue of Dead Reckoning, a new weekly newsletter written by me (Nick Graham), founder of Aptum AI.
You’re receiving it because we’ve crossed paths in business, the military, a gym, a church, or a bar, and you seemed like the kind of person this is for. I’ll be writing more specifically for founders and operators at PE-backed platforms trying to make better decisions in noisy environments, but even if that isn’t you, exactly, you might still find it useful. Perhaps even interesting or enjoyable.
If that is not you, or you do not want this in your inbox each week, you can unsubscribe below.
Each week, I’ll share notes on strategy, data, systems, leverage, and decision-making under uncertainty.
The title of the newsletter comes from land navigation. For my military friends, pardon the explanation below (Air Force excepted, this will probably be new).
In difficult terrain, you do not move well by observation alone, and you do not move well by instruments alone. You use both. You take a bearing and count your paces. You look for handrails and backstops. You adjust continuously.
That is how I think businesses should operate too, especially now.
We are living through a period where companies have more dashboards, more data, more software, more automation, and more AI than ever before. In theory, that should make decision-making easier. In practice, it often creates a different problem: teams become surrounded by tools while losing contact with reality.
That tension is what this newsletter is about.
You can expect one issue each week.
When the map and the terrain disagree
In business, tools and intuition are supposed to reinforce each other.
Your dashboards, scorecards, CRM stages, forecasts, and models should generally line up with what experienced operators are seeing on the ground. When that happens, the business feels coherent. The instruments make sense. The situation is legible.
But sometimes they don’t match up.
Most people have been in a meeting where the customer health dashboard says an account is fine, while the account manager says he has a bad feeling.
Or, where a forecast says the quarter is on track, but a sales leader doesn’t feel so confident.
Lately, this is one I see more and more: an engineering lead says an “AI-enabled” workflow is performing well, while the people who are supposed to be using it keep routing around it, copy/pasting into Claude or ChatGPT.
When the map and the terrain disagree, which one should you trust?
Dead reckoning and terrain association
In land navigation, there are broadly two ways to move.
The first method is dead reckoning.
You plot your distance and direction on the map, take a bearing, and move deliberately while counting your paces to track progress. It is structured, measurable, and disciplined. It gives you a way to move when the terrain is ambiguous and lacks easily identifiable landmarks.
The second method is terrain association.
Instead of navigating primarily by calculated bearings and pace counts, you use visible terrain features: a river, a road, a ridgeline, a hilltop, a treeline. You orient by what is obvious and observable in the world around you.
Good navigators do not treat these as competing methods. They use both.The map and compass helps you move with discipline, and the terrain lets you know you’re moving in the right direction.
Business works the same way. Or, at least it can.
What this looks like in business
Dead reckoning in business looks like:
- Dashboards with KPIs
- Forecasts & Predictions
- Structured Workflows & Automations
- Pipeline Stages & Playbooks
- Customer Health Scoring Models
Terrain association looks like:
- Direct customer conversations
- Operator judgment
- Frontline observations
- Repeated objections
- Subtle signs that a team, process, or account is drifting
Neither is enough on its own.
If you rely too heavily on the compass and your pace-count, your business can become organized, elegant… and completely detached from reality.
If you rely too heavily on the terrain, you may stay close to reality but struggle to scale, repeat, or compound what you know. Good businesses must learn to do both.
They learn to use structure without becoming trapped by abstraction. They learn to stay grounded in reality without becoming purely reactive.
A concrete example: NNACV prediction
At ServiceNow, I worked on a machine learning prediction for one of the company’s top KPI’s: net-new annual contract value (NNACV). The prediction refreshed every hour and appeared on the CEO dashboard alongside 9 other KPI’s he checked every morning.
It was built on 3 years of historical CRM pipeline data and predicted how much NNACV the company would close in the current quarter based on factors like total pipeline, weighted pipeline, and historical conversion rates. All of this was calibrated by region, product mix, customer segment, and many other factors.
The sales team’s quotas, goals, and commissions were all tied to NNACV. NNACV was, and still is, a huge driver of the company’s operating rhythm.
The ML NNACV prediction showed up alongside the “Manger Expect” number. It was a perfect “human vs. machine”, “map vs. terrain” comparison.
I was regularly called into high-level meetings to “defend” the ML prediction against what a sales manager was saying. Sometimes they thought it was too optimistic, and sometimes they thought it was too conservative.
Who ended up being right more often: the ML model or the manager? The answer I began to observe, anecdotally, was this: whoever had the more pessimistic view was right more often.
If the ML model predicted a team would hit 102% of quota, but the manager said “no, we’re going to miss this quarter,” the manager was right more often than not.
If the ML model predicted a team would only hit 88% of the quota, but the manager said “we’re going to hit quota, trust me,” the model tended to be right.
From these discussions, we eventually arrived at a simple approach that actually increased sales performance. If the model or the manager flags a risk of missing quota for a particular region, it gets in-depth scrutiny and analysis from the sales and the ML team. That analysis often produced an all-hands-on–deck effort to close the gap.
The deeper lesson
An ML prediction is not the terrain itself, but the terrain is often too noisy to navigate at scale without tools.
That is the deeper lesson, in my opinion.
The point is not to choose between data and judgment; the point is to build a tighter loop between them.
That can apply to forecasting, pipeline management, hiring, product analytics, internal workflows, and operating reviews.
It applies especially in the kinds of companies I spend time around: PE-backed companies and platforms trying to professionalize operations without suffocating them.
Most of these companies are founder-led and family owned, and they’ve spent so much time walking the terrain of their business that they don’t need a map or a compass. They know everything by heart, and they and their employees run the business by habit.
But, the value creation plan demands mining the current terrain for untapped potential, moving into entirely new terrain, or both. It requires codifying the “gut feel” of the founder in data, tools, and systems that scale, but without losing human intuition.
Why this matters even more with AI
This is an incredible challenge, and not everyone is up to the task. For those who are up to it, and for those who learn to do it well, the upside is immense.
Paradoxically, AI makes the challenge harder, not easier. It will also create greater divergence in outcomes based on how well it’s used. Some businesses will ride the wave, and others will get crushed by it.
AI systems are powerful forms of dead reckoning. They classify, score, summarize, predict, and recommend. They turn messy reality into a cleaner abstraction that can be acted on more quickly. That is part of its value. It is also where the risk enters.
AI can create the illusion of understanding long before it has earned it. AI applied across messy, inconsistent data will produce the same “confident” answers as AI applied to clean, well–structured data with a strong semantic layer. But, that’s like navigating with a compass that isn’t properly calibrated, or using a pace-count set in a level, open field while walking up a steep hill full of obstacles.
The results may look sophisticated, but eventually you’ll look around and see terrain features that don’t match where you thought you were.
AI is useful. When it is grounded in a strong data substrate, it is sometimes decisive. An AI or ML model isn’t “set it and forget it.” The NNACV ML was more useful with sales manager feedback.
In other words: AI tools are only useful when they stay connected to the terrain.
Why I’m writing this
This is the territory I spend most of my time in through Aptum AI.
The surface-level version of the work is building data-powered systems that improve both speed and accuracy, and that enable AI-powered processes, and operating layers, that help companies move with more clarity.
At a higher level, this work is helping a business create a better map without losing contact with the reality of the terrain.
Most business processes fail in one of two ways:
- They are all terrain: dependent on tribal knowledge, improvisation, and operator instinct that cannot scale.
- Or they are all map: elegant, well-structured, and detached from how the business actually works.
The right answer is usually some disciplined combination of both. That is what I want to explore in this newsletter.
Not AI for its own sake or dashboards for show, and not abstract strategy disconnected from execution.
I want to explore a more practical question, examined from different angles each week: How do good operators navigate uncertain terrain well?
Closing
That is the idea behind Dead Reckoning.
Each week, I’ll use this newsletter to explore strategy, systems, AI, leverage, and decision-making under uncertainty, with one eye on the map and one eye on the terrain.
If you suspect your map and terrain have drifted apart (or you're a Marine and your map is drawn in crayon on the back of a napkin) reply and tell me where the mismatch is showing up.