The Visibility Problem
Google's AI is remarkable at optimising toward measurable signals. Conversions. Revenue. ROAS. Give it a target and clean data, and it will find ways to hit that target.
But here's what it can't see:
- Your actual margin on that order
- The return rate on that product category
- Your cash position this month
- The inventory you need to clear before it becomes a liability
- The LTV of different customer segments
- The credit terms you negotiated with suppliers
These aren't obscure edge cases. They're the variables that determine whether your business is actually profitable.
Why This Gap Widens
As AI gets better at optimising visible metrics, the invisible metrics become more consequential. Here's why:
AI drives convergence. When every advertiser uses the same automation, everyone optimises toward similar outcomes. The only remaining differentiation comes from the factors the AI can't access.
Complexity compounds. Modern ecommerce involves more moving parts than ever—multiple sales channels, complex supply chains, varied payment terms. The P&L keeps getting more nuanced while the AI stays focused on simple conversion metrics.
Economic pressure increases. When margins are thin, the difference between a 25% gross margin product and a 35% gross margin product is everything. The algorithm treats them the same.
Forecasting, Cash Flow, Stock Risk
Let's get specific about what falls through the gaps:
Forecasting blind spots. AI optimises based on historical patterns. It doesn't know that you're planning a price increase next month, or that a key supplier is about to raise costs, or that your biggest competitor is exiting the market.
Cash flow timing. A sale today might not become cash for 30-60 days after returns and chargebacks. The algorithm celebrates the conversion immediately. Your accountant sees a different picture.
Stock risk asymmetry. Pushing hard on low-stock products creates stockout risk. Ignoring high-stock products creates markdown risk. Neither is visible in the ad platform.
The Human Layer
This is why human judgment isn't going away—it's becoming more important.
The role of a good operator isn't to out-optimise the machine. It's to feed the machine with commercial context it can't access on its own. To override when the algorithm's blindness creates real business risk.
That means:
- Adjusting targets based on margin, not just revenue
- Pulling back when cash needs dictate, regardless of ROAS
- Prioritising products based on business need, not conversion potential
- Making decisions the algorithm would never recommend
Building Visibility Bridges
The sophisticated approach is to bridge the gap—to create systems that incorporate P&L reality into campaign decisions.
This might mean:
- Margin-based bidding that adjusts targets by product profitability
- Inventory feeds that dampen spend on low-stock items
- Custom segments that account for customer LTV differences
- Governance rules that cap spend based on cash position
None of this is automatic. All of it requires understanding the business, not just the platform.
The Enduring Advantage
The gap between what AI sees and what your business needs will keep growing. That gap is where value lives.
Agencies that understand this invest in commercial fluency. They learn to read P&Ls. They ask about cash flow. They treat campaign management as business management, not platform management.
The ones that don't will keep celebrating ROAS while their clients wonder why profitability keeps shrinking.