When a GTM team decides to “implement AI,” they usually start by trying to automate the most visible thing a rep does: writing emails.
This is almost always the wrong place to start.
Automating the final visible output (the message) before you fix the underlying input (account research, signal routing) is a recipe for scaling noise. If you want real commercial execution speed, you start where the work is repetitive, high-frequency, and low-judgment.
The Problem With Visibility Bias
We automate visible tasks because they feel like they save the most time. If a rep spends two hours a day writing emails, automating that feels like a massive win. But you haven’t actually removed a bottleneck—you’ve just moved the friction to the prospect’s inbox by sending lower-quality, hallucinatory emails at volume.
You don’t want to automate the part of the job that requires judgment and empathy first. You want to automate the parts of the job that require zero judgment.
Start With Latency and Consistency
The best first automations in a sales or marketing system do two things:
- Reduce Latency: How long does it take for a signal to become actionable?
- Increase Consistency: Are we enriching the same baseline data against every account without exception?
Here is the hierarchy I use when deciding where to drop an LLM or an API-driven workflow into a new team.
Step 1: Automate Data Enrichment
Instead of asking AI to write a message, ask it to read a 10-K and extract the three strategic priorities of the company. Ask it to read a LinkedIn profile and summarize the prospect’s last three jobs. This requires zero creative judgment. It simply structures unstructured data so your human reps don’t have to read PDFs for 45 minutes before a cold call.
Step 2: Automate Routing Logic
If you have 50 inbound leads an hour, and an MDR is clicking through Salesforce to figure out who gets what based on industry and employee size, you have a latency problem. Use deterministic logic to hit an API like Clearbit or Clay, append the firmographics, and route the lead instantly.
Step 3: Automate Context Assembly
When a lead is passed from a BDR to an AE, the system should instantly compile a one-pager into the CRM notes: Here is the company, here is the person, here is the signal that triggered them, here is the original inbound request. The AE should never have to piece the story together across three different browser tabs.
The Takeaway
Automate for system reliability, not novelty. Automate the things that nobody wants to do anyway. Once the data flows cleanly, the routing is instant, and the context is assembled automatically—then you can start letting AI draft the messaging.