# A System of Action

*Published 2026-10-16* | Author: brad-webb

> **tl;dr** More truth than ever. Less time to act on it. Every tool in a growth stack is a source of truth now, and the best it has ever been. The action still lives with a person, on a hunch. What's worked for us is a system of action that sits outside the records, reads all of them, runs rules we wrote, and keeps a receipt for every move. I gave this as five minutes at The State of B2B GTM Summit on Wednesday; this is the longer version.

*The five minutes were in the founder-led block at [The State of B2B GTM Summit](https://stateofb2bgtm.com/). GTM is the sales and marketing side of the company. Monday's note has [the coverage](https://www.civic.com/field-notes/whats-working-in-b2b-gtm). Here are the same seven beats, each given the room a five-minute slot can't.*

## More truth than ever

I run growth at Civic. What's working for us right now, and what I'd suggest to anyone in this space, is a system of action.

The frame fits in one line. More truth than ever. Less time to act on it.

We've all got more records than we've ever had. More events, more sources of truth, and more of it accurate, because the tools got good. And we've got less time to do the most with any of it, because the tools also got good at filling up faster than a person can read. That gap is the whole talk. On stage I had five minutes for it. Here I can take the beats one at a time.

## The records

I've seen a lot of stacks. Here's a good one. A CRM. Product analytics. A warehouse, maybe. Something that tells you which companies are on your site. Data and enrichment (what fills in who a visitor is). A sequencer, or three (the tool that sends your email steps). Campaigns aimed at named accounts. Forms, and whatever routes a form to a person. A CMS. Nine categories, and I've left the vendor names out on purpose, because it doesn't matter which ones you picked. They're all good to great at what they do. Every one is a source of truth. It's the best that stack has ever been.

Here's the catch. A team of one acts in one tool at a time, and usually on a hunch or a bet. The other systems don't stop. They keep filling up with things worth acting on, and whichever one a person feels most at home in gets acted on. The rest wait their turn, and the turn comes late or it comes never.

So, lots of systems of record. The action still lives with a person, who still has to do something, even if that something is prompting an agent to work out what to do.

THE RECORDS (interactive: tick the systems of record you run, then say where the action lives today)
Nine systems of record, and what each one is the truth about:
- CRM: accounts, deals, owners
- Product analytics: events, usage, activation
- Data warehouse: history, models, cohorts
- Visitor identification: which companies are on your site
- Data and enrichment: firmographics, contacts, intent
- Sequencing: sales engagement, outbound
- ABM and campaigns: account-based marketing, ads
- Inbound and forms: form capture, lead routing
- CMS and content: pages, what they read
Each one is a source of truth. None of them acts.
Where does the action live today? A person, on a hunch. A person, on a schedule. Or a rule we wrote.
Figure 1. The nine records, and where the action lives today. Tick the ones you run; the line under them says what they have in common.

## Why we built one

Bryn is the agent we built for this. A rule you set once, when this happens, do that, which Bryn calls a Play. It sits outside the records, reads all of them, scores what it sees, runs the Play you approved, and writes the receipt. We were its first customer, and a captive one.

That was us. More systems with more data than ever, and at some point we realized we needed something that could act on it while it was still worth acting on. So we built one. Looking back, we were searching for upside, watching for downside, and trying to keep ourselves honest. And sane.

Upside first. The next customer might look nothing like the last one. As our own first customer, we had no idea which door the next one would walk in through. The site? The product? Somebody replying to an email? An intro? A note like this one?

Downside. We wanted to keep the most important customer anyone has: the one already paying. A signal is just something a buyer did, like going quiet in a trial, or coming back to your pricing page. The quiet trial lives in the same records as the new buyer showing up. Both deserve a reader. A small team has one pair of eyes, and it feels churn harder than most.

And honest. The janitorial work is where the details hide. The repeat visit. The second person from the same company. Somebody has to read all of it, every day, and a founder's day doesn't have that hour. It shouldn't, either. No judgement. Mine didn't have it.

SEARCHING FOR UPSIDE, WATCHING FOR DOWNSIDE, KEEPING OURSELVES HONEST (interactive: pick one of three reasons; the panel gives the reason and one example)
1. Upside. The next customer might look nothing like the last one. As our own first customer, we had no idea which door the next one would walk in through. Example (illustrative): a new buyer at a door you did not expect, replying to a note like this one with a question about pricing.
2. Downside. We wanted to keep the most important customer anyone has: the one already paying. A trial going quiet lives in the same records as a new buyer showing up. Both deserve a reader. Example (illustrative): a trial going quiet, a busy first week and then a silent one.
3. Honest. The janitorial work is where the details hide. Somebody has to read all of it, every day, and a founder's day doesn't have that hour. Example (illustrative): the second person from the same company, on the pricing page an hour after the first.
Figure 2. The three reasons, taken one at a time. Pick one; the panel gives the reason and one example of the kind of thing it catches (illustrative).

## Outside, and open

So here's our take. The system of action should sit outside your systems of record, watching all of them, so each one gets to stay the best at its job. Every event lands in one place and gets a score, and that's what makes acting reliable: the score reads the event, the account, and the state the account is in. We went the open route. Our stack stays where it is, and it keeps changing, which is how stacks are.

Open means two plain things. The records keep their jobs, and the agent keeps the receipts. Your CRM stays your CRM. Bryn reads the record and writes back to it, and the record stays the record. The product, the site and the warehouse each know something the CRM doesn't, and outside is the only place that reads all of them at once. If someone asks why we didn't build this inside the CRM, that's the answer: inside, the CRM is the only truth that counts.

The other side of being outside is what you can see from there. Third-party signals, who raised, who's hiring, who's shopping, are good starting points, and you can buy or rent them. First-party signals, what somebody actually does on your site, in your product, in your emails, are the signal you can't buy. Only you can see it, measure it, route it. That's what closes deals. Hold it at arm's length and it costs deals.

Watch, score, act, log. And yes, a system of action should have its own system of record. Every run is logged, so the people who sign off can read it.

OUTSIDE THE RECORDS. OPEN TO ALL OF THEM. (static diagram, illustrative)
Starting points (third-party, rented): Enrichment (who raised, who's hiring); Intent data (who's shopping).
The signal you can't buy (first-party, yours): Your site (pricing page, again); Your product (a trial going quiet); Your CRM (owner, stage, notes); Your funnel and sequences; Your forms and content.
Bryn by Civic, the agent: Watch (every record), Score (the state), Act (your rule), Log (the receipt). Every event, from every source, scored: the event, the account, the state. Act on what buyers actually do, the minute they do it. Outside: reads every record, writes back. Open: your stack stays where it is. Every run is logged.
Where each run lands: CRM, task and note. Owner, ping, with context. Sequencer, enroll. Discovery, book the call. Buyer, the follow-up you approved. Record, its own system of record. Somewhere a person can read it.
Figure 3. Outside the records, open to all of them: starting points you can rent on the left, the signal you can't buy beside them, the four steps in the middle, and where each run lands on the right (illustrative).

## A short confession

We started Bryn in Approve mode, because I didn't trust it yet. Approve mode: Bryn drafts the step and waits for your yes. I read every single one. Yes. Yes. Yes.

After a few weeks the yeses got boring. Which, it turns out, is exactly the signal you're waiting for. I told the longer version in [a piece I wrote about being our own first customer](https://www.civic.com/field-notes/we-are-customer-zero), including the draft that was wrong and showed us a rule we'd written too loose. We tightened the rule. The log still has the day it changed.

If you do one thing Monday: turn Approve mode on, read every yes, and note the day they get boring.

A FEW WEEKS OF APPROVE MODE, ONE YES AT A TIME (interactive: sixteen slots; each tap says yes to the next draft; illustrative)
We started in Approve mode: Bryn drafts the step and waits for your yes. Yes. Yes. Yes. One draft in the run was wrong, and it showed us a rule written too loose; we tightened the rule and the log kept the day. After a few weeks the yeses got boring. That is the signal: move the rules you trust to run on their own, and keep the rest on Approve. (illustrative; a few weeks of Approve mode)
Figure 4. A few weeks of Approve mode, one yes at a time (illustrative). Keep going until the line changes.

## The tap on the shoulder

This is the part I actually care about.

Founders work best in tunnel vision. That's how anything gets built. So an agent that reads everything you don't have time to read is a tap on the shoulder. Take a breath. Take a beat. Look at this one.

A tap is small on purpose. It's an account, a line and a reason, at the moment it's worth a look. Then you go back to building. The reading happened somewhere else, all day, and you only hear about the part that matters.

Because the most important signal you'll ever catch is the one that proves you wrong. The company you'd written off, back for the third time. The one you were sure about, that never came back. The draft Bryn got wrong, which showed us the rule we'd written too loose. You only get those when something is reading every record, every day, and pushing against the rules you've bet on.

Chris, our CEO, wrote on Wednesday about [the buyer whose AI does the research](https://www.civic.com/field-notes/your-buyers-ai-does-the-research) before a person ever shows up to verify it. The return visit and the second person from the same company are that person showing up. They land in the janitorial work, the part of the day nobody has. A tap on the shoulder is the only way I know to get them read.

## A bet for your 2027

Everyone who spoke in my block at the summit is making it easier to know who's interested. What's worked for us is what happens next.

So that's my bet for your 2027. Keep your systems of record. Put a system of action on top. Outside, open to all of them, running rules you wrote, with a receipt for every move.

If you take one thing from the five minutes, take the order. Records first, because they're the truth. Action on top, because the scarce thing is what happens next. Budgets for the year ahead get set in the next few weeks, and most of them are still written as if seeing were the hard part.

Where Bryn sits, in one paragraph. Outside your systems of record, open to all of them. Bryn is not another dashboard to watch. It is the governed execution layer that runs Plays through your stack. In plain terms: it reads the nine records above and runs the rule you approved.

I'm Brad, at Civic. Bryn lives at [civic.com](https://www.civic.com/).

**Stop watching signals. Start running them.**

Source: https://www.civic.com/field-notes/a-system-of-action