Every tool your agency pays for now has its own AI, and none of them can see the others. What tames the Franken-stack is a layer that sits across all of them.
Recently, we talked with an agency that is rolling out a new project management system. Their operations lead likes the new system. She also said, “There’s so much that I don’t think that system can do.” Her examples were creating order confirmations and warning the team before a client runs out of hours.
Neither of those is a project management problem. An order confirmation needs the signed scope, the rate card and the client’s billing contact. A time bank warning needs the contract, the hours logged and the work still on the schedule. The answers sit in four or five different tools, and no single tool can see all of them.
Most agencies run on a tech-stack like this one: a CRM, a project tool, a notetaker and a finance system, each with its own AI. What’s missing is one layer across them that brings the answers together.
Count the AIs
Make a list of the software your agency pays for and check which ones added AI this year. It is probably all of them.
Your CRM has agents. ClickUp has Brain, and this month Atlassian announced it is bringing OpenAI’s newest models into Jira and Confluence. Notion has agents. Your meeting notetaker writes summaries, your reporting platform writes insights, and your proposal tool scores RFPs.
Each of these is useful. The AI in your CRM is good at your CRM. The notetaker is good at meetings.
Each one also knows only its own slice. The CRM’s AI knows the pipeline, and it has no idea the account team is over capacity. The project tool’s AI knows the schedule, and it has no idea the client’s contract renews in four months. The notetaker heard the client say the last campaign felt off, and it can’t tell you what that client pays or who on your team needs to hear it.
Agencies already live with this. Someone senior carries the whole picture in their head and connects the pieces by hand. Adding AI to each tool makes every slice faster. The connecting stays exactly where it was: with that person.
Agencies have built this for their clients
If a knowledge layer sounds abstract, look at what agencies have been building for brands for the past decade.
Brands hit this wall with their customer data. The website ran on a content management system. Email ran on another platform, ads on another, analytics on another, commerce on another. Every tool knew something about the customer, and none of them knew the customer. The industry’s answer was the digital experience platform, with a customer data platform underneath it: one place that holds a single view of the customer across every channel.
A good DXP connected to the email platform and the ad tools and gave all of them a shared understanding of the customer. The tools stayed. What changed was that they finally worked from the same picture.
Plenty of agencies have scoped, sold and implemented those platforms. The knowledge layer is the same move pointed inward. A DXP gives a brand one view of its customer. Signal78 gives an agency one view of itself.
What the layer holds
Signal78 keeps your CRM, your project management tool and your finance system where they are. It connects to them, reads across them, and holds two things none of them were built to hold.
The first is how your agency runs. Your clients and what each one needs, your people and what they’re good at, what you’ve charged, why you won and why you lost, how your agency writes, and the calls your senior people make that nobody has written down. We call that the brain.
The second is how agencies run in general. While building Signal78, we wrote in a layer of agency practice: how value-based and hourly pricing each work and where each one breaks down, when renewal planning should start, and the common ways scope creeps in before anyone notices. Your agency’s own patterns sit on top of that and adjust it. If your team keeps saying a person is over capacity, the system stops suggesting them for new work.
Put the two together and the questions that cross tools get answers. Which clients are about to run out of hours this month. Which renewals are six months out, and who should start planning them. What the new-business team promised that client services now has to deliver.
See it work
Should you build it yourself?
Some agencies will build this layer themselves, and that’s fine. One capable person with AI can get a convincing version working in a weekend.
The weekend is the easy part. What follows is a permanent job. Someone has to keep the connections working as every tool changes its AI, decide what the system reads on each job, keep each client’s material separate from the next client’s, and keep the whole thing running after the person who built it leaves. That is software maintenance. Most agencies are not in that business, and they shouldn’t be.
If you’re considering building it yourself, our field guide lays out what it takes to keep an agency AI system current, secure and separated by client, so you can decide whether it’s a job to take on yourself or one to trust to a platform like Signal78.
At the end of our conversation, the operations lead, she put it better than we could. “[Signal78] takes AI and makes it a little less scary, gives you a partner,” she said, “as opposed to me trying to figure this out myself. Or hiring people in-house to figure it out.”
That is the job we’ve taken on, so agency people can spend their time on the strategy and creative work they got into this business to do.
Teri Sun
A 20-year agency leader and innovation driver, with deep expertise in putting AI to work across complex agency teams. Teri leads how Signal78 is built, shaped and brought to agencies.
