In 1988, aged 18, I arrived at a field artillery battery in Baumholder, West Germany, in the last years of the Cold War. A battery exists for one purpose: to put 155mm rounds on targets its gun crews will never see.
My role was logistics, not gunnery. But I watched how the system worked, and almost four decades later it is the clearest model I know for a question many executive teams are now asking privately. We have bought the capability. Why isn't it landing?
The guns never saw the target
A 155mm howitzer of that era could reach about 18 kilometres with conventional rounds.2 The crews knew their guns, their ammunition, their rates of fire and their tolerances. They could deliver precise effects at ranges far beyond anything they could see.
What they lacked was line of sight. From the gun line, the battlefield was a set of coordinates. The battery held enormous capability and no view of where to apply it.
That was not a flaw. It was the design. Capability sat at the rear, where it could be maintained, supplied and massed. Sight went forward, where the fight was. Everything depended on what connected the two.
The observers never fired a round
Forward observers belonged to the artillery. They lived with the infantry battalions we supported, alongside the companies at the front.
They saw what the guns could not. Just as important, they knew what the guns could do: ranges, effects, response times, what a battery could deliver and what it could not. That knowledge is what made their sight useful.
When an observer found a target, they sent a call for fire. It followed a fixed sequence: who is calling, what kind of mission, where the target is, what it is, and how to engage it.1 The fire direction centre turned the call into firing data. The gun crews set charge and elevation, and fired. The observer watched the rounds land and sent corrections.
How did sight at the front reach capability at the rear?
Adjust. The observer watches where the rounds land and sends corrections until the next volley can fire for effect.
No one in that chain did anyone else's job. Each understood the next link well enough to hand over cleanly. Most enterprises have built one end of that chain and not the other.
Most enterprises now have the guns
Model access, platforms, central data and AI teams. Enormous capability, and little line of sight to where it should land.
The lines of business see targets every day. The claims leader who knows exactly where the process breaks. The credit manager who can name the decisions that take too long. The maintenance planner who knows which failures repeat. They have sight. What they lack is fluency in the capability behind them.
The industry's answer is the forward deployed engineer. Palantir made the role famous, and its name borrows the military's language. AI labs are now building the same function: OpenAI set up a forward deployed engineering team in 2024, and Anthropic is growing its applied AI group.3 These are rare people who combine engineering skill, domain knowledge and applied AI experience, embedded with the customer.
The model works for the vendors. Putting capability next to the problem closes the same gap the artillery closed. But most enterprises draw the wrong conclusion from it: that AI readiness is a hiring problem, solved by bidding for a scarce corps of outsiders.
AI readiness is not a hiring problem. The people who can see the targets already work for you.
Your forward observers already work for you
The people who can see the targets are already in your building. They run your claims teams, your credit books, your supply chains and your maintenance schedules. They have domain depth no outside engineer will build in a six month engagement, and they see the friction daily.
What they lack is the observer's kit. It has three parts.
- FluencyKnowing what the organisation's AI capability can and cannot do: its ranges and its effects.
- A call for fireA disciplined way to describe a target, the context around it, the data available and the effect required.
- A fire direction centreA small applied team behind them that turns a well formed call into something that works.
The Army did not ask infantry commanders to master gunnery, and it did not put gun crews in foxholes. It trained a small corps of observers, attached them to the artillery and placed them with the infantry. The arrangement was deliberate.
Enterprises can build the same structure with people they already have. Not every domain expert. The ones with the pattern the Army selected for: people who know their ground, can describe a target precisely and can be trusted with the call. Which raises the obvious worry.
Won't this multiply ungoverned pilots?
The strongest objection is control. Give hundreds of domain experts a way to call on AI and you get hundreds of pilots, duplicated spend and risk nobody approved.
The artillery faced the same problem with higher stakes. Its answer was not to take the radio away from the observer. It was structure. Every call went through the fire direction centre. Fires were checked before they were delivered, so rounds did not land on your own side. The observer held the authority to call. The system held the authority to fire.
The enterprise version works the same way. Observers call. The applied team checks each call against the data, the risk and what already exists, and then builds. Control sits in the fire direction centre, where it adds speed, not in a committee the call must wait for. That structure is what turns individual observers into a system.
Readiness is a system, not a weapon
Most AI investment still flows to the guns. Bigger models, better platforms, larger central teams. But the guns improve on the vendor's schedule whether you act or not. What differentiates is the observer corps and the system that connects it to the capability.
The kit makes an observer. Two further conditions make a system. A shared language, so a call for fire means the same thing to the business, the applied team and the platform. And authority, so an observer can call a mission without waiting for a steering committee.
What connects the people who see the problems to the capability that can solve them?
Ranging round. A pilot that misses is corrected, not cancelled.
A system built this way still misses. What matters is what it does next.
First rounds rarely land
The artillery took one habit for granted: adjustment. When the target location was uncertain, the observer did not fire for effect. They called for adjustment. The first round was a ranging round. The observer watched it land, corrected, and only then fired for effect.1
Most enterprises treat a pilot that misses as a failure. A battery treats it as a ranging round, provided someone is watching where it lands and someone adjusts. The failure is not the miss. It is firing with no observer, or observing with no way to correct.
The call is yours to make
Advantage will go to the organisations that develop the observers they already employ: domain experts who know what the guns can do, have a disciplined way to call for fire, and hold the authority to make the call.
You don't hire that in. You build it.