We don’t know what we don’t know, and our assumptions can bury us. This is an extraordinarily great example of what can happen (at a different scale) in human relationships, marriages, child rearing, career path, finances, economies, politics, war…
What could go wrong if there was perfect communication and AI-enhanced merged lookaheads?
Well let’s see, the attached example of the data center. What could go wrong? Even with perfect communication? Reporting honesty, hidden competency and skill level damage, material science supplier QC, weather or logistic anomalies, and, generally, the second law of thermodynamics, entropy.
But clearly, of course, we should use what tools we have available—to do better. We can.
__________
Pejman Golkar
I have 27 subs building a 500MW AI data center. None of them know what the other 26 are doing.
This is one of the most expensive problems in data center construction. Nobody talks about it.
Every sub has a plan. Electrical has a plan. Mechanical has a plan. Civil, structural, fire protection, controls, commissioning — they all have plans.
None of those plans talk to each other.
The GC builds a master schedule in P6. It looks perfect. Every activity sequenced. Every dependency mapped. Every milestone dated.
Then the job starts:
In the field, each sub runs their own sequence based on their own reality:
→ Labor availability changes weekly → Material deliveries slip without notice → Equipment access gets blocked by another trade → Permit timing shifts the start date
Sub A finishes early. Sub B isn’t ready. That gap doesn’t show up in P6. It shows up as 40 workers standing around waiting for a clear area that won’t be clear for 2 more weeks.
On a 500MW+ build, that gap costs $150K–$250K per week in extended general conditions alone.
Here’s what makes it invisible:
Every sub submits a 3-week lookahead. That’s 27 lookaheads landing on the GC’s desk. Each one built independently. Each one assuming the other 26 are on track.
Nobody is cross-referencing 27 lookaheads against the master schedule, material deliveries, and equipment access — every day, across every area.
No human can. The volume is too high. The connections are too many.
So what actually happens?
The superintendent walks the site. Sees a problem. Makes a call. Moves a crew. Reacts.
That works at 50MW. It breaks at 500MW.
At 500MW you have multiple buildings, multiple shifts, multiple phases running simultaneously. The super can’t walk every area every day. The weekly coordination meeting is outdated by Tuesday afternoon.
The information to prevent these conflicts already exists. It’s in the lookaheads, procurement logs, vendor updates, RFI tracker, submittal register.
It’s all there. It’s just not connected.
Here’s what connected looks like:
→ Sub A’s lookahead shows they’re finishing Area 3 Wednesday → Sub B’s lookahead shows they’re starting Area 3 Monday — 5 days early → Procurement log shows Sub B’s material delivery slipped to Friday → Connected: Sub B can’t start Monday. Sub A has 2 extra days of float. No conflict. No standby crew. No $200K week.
That cross-reference took 30 seconds.
Doing it manually across 27 subs, 50 areas, and 10,000 procurement items takes a full-time person who doesn’t exist on most projects.
This isn’t a people failure. These are the best teams in the industry managing impossible complexity with tools that weren’t designed for this scale.
The answer isn’t more meetings. It isn’t a better spreadsheet. It’s connecting what your systems already know so the conflict gets caught on day 2, not week 6 when the crew is on standby.
How many subs are on your site right now? How many of their lookaheads have you cross-referenced this week?
