Cutting Workload by 60% with Agentic Workflows

With every hour used up, an agentic workflow gave us a few of them back.

Dale has a problem.

I should know, I'm his business partner.

He has 3 hours of time but 8 hours of work. Sound familiar?

To make the problem worse, it's work that's easy for a small business like ours to ignore: Business development.

A couple months ago we started Latent to help deep tech companies surface their brand to their market. But Latent is new, people don't know us yet.

At a basic level, we want to go out and:

• Find the deep tech companies where we can add value

• Talk to the right folks at these companies

• Keep track of it all

The Agentic Workflow Solution

Here's how I approached the problem with an agentic workflow:

1. Define the boundaries: For example, when we say Deep Tech, who do we mean? What are the segments where we can add value? Where are we keeping track of it? Where can we find the information we need?

2. Daily monitoring: Have an agent scan defined news sources for signals from companies matching our segment profile. It runs the company through a scoring system which scores on fit and relevance, checks our CRM to see if we already know them, and looks at all past context if we do.

3. Weekly round-up. Every Monday morning Dale and I get an email spelling out who's worth talking to and why. We give the agent the go-ahead for the right companies and it assigns us tasks in our CRM.

4. We do the work. We complete the tasks (things like send an email, connect on LinkedIn, arrange to meet at an event, book a meeting, etc.). This is where 99% of our time is now spent.

5. Ongoing follow-up. our agent reads our tasks for status updates and helps with next steps, e.g. it suggests what to put in a draft outreach message, sets a reminder if someone says "come back next month", writes a research brief before meetings, or makes a note if someone shouldn't be bothered again.

6. Self-review. Every few weeks a self-assessment automated process runs to evaluate system performance and reports on conversion rates through each stage, so we can see what needs improving.

Before: ~8 hrs/week on research, admin, CRM updates, outreach, follow up, etc.
Now: ~3 hrs/week, spent on the human part: who to talk to, how to reach them, and what to say (after all, we want to be helpful, not annoying).

The time spent upfront for setup is meaningful, but the time payback compounds every week.

Why Systems Thinking is the Real Opportunity

My favourite part of building solutions like this is not solving the problem. It's that AI has created so many options as to how solve the problem that the challenge has shifted from technical implementation, to judgement and systems thinking.

• Sometimes AI's approach to parts of the workflow is just weird.

• Sometimes it's not secure enough.

• Sometimes it creates downstream impacts which only show up later.

The realm of what's possible has become so huge. I could do all of this with a series of prompts, scheduled tasks, custom GPTs, managed agents, custom coded applications, traditional workflow orchestration... or a combination of those.

Understanding the system and the bounds within which I have to operate, like cost, compliance, security, means that not only can we save 5 hours a week but we can do it in a way that scales with us... or can be translated and built for our clients.

Also, as a bit of a bonus, it hasn't cost us any additional money in software licenses.

Have I solved all of Dale's problems? I'm not a miracle worker.... but I have saved him 5 hours every week.

---

Simon Jared is Co-Founder of Latent, a B2B marketing consultancy for deep tech. He helps companies build AI-first marketing operations.

Book a call.