This is a post about building, learning, and loving our new homegrown CRM and knowledge base. Like other go-to-market teams, we use a broad range of tools, but we'll focus just on these two for the purposes of this post. This is the first in a series we'll write about building sovereign, agentically-enabled workflows for our sales and marketing teams so we can collaborate more effectively and do better work.
A year ago, our sales and marketing team was using a big, well-known CRM, and our institutional knowledge was spread across a number of different places: call transcripts, decks, materials in Google Drive, Slack, and more.
12 months later I've just returned from maternity leave to a new operational environment.
Sharing how this works is important for me. We believe strongly in sharing our learnings with the community, and we are increasingly making this recommendation to clients: that the systems holding their most business-critical information should be sovereign (owned and operated by them), not rented from someone whose priorities can change. That's a significant thing to ask a company to take on. It would be a thin recommendation if we weren't doing it ourselves.
So this is us walking the talk.
Why ownership and AI are the same conversation
Those two threads, becoming agentically enabled (learning to work with AI agents) and owning our systems, look like separate projects. They aren't, and the connection is the reason we did any of this. Sovereign technology is an accelerant. Three reasons, and they all come down to speed and control. Which really just means we get to do better work.
You can change it as fast as you learn. When a team is figuring out how to work with agents, the requirements change every week, because working at the cusp, you genuinely don't know in advance what you'll need. Changes to our own CRM and knowledge base happen as we need them, usually in hours or days – not through a feature request, a roadmap or a quarter. A system you can't change at the speed you're learning is a brake on the learning.
An agent is only as useful as what it can reach. Connecting our agents to systems we own was a decision we made and then made happen. Doing the same with a platform you rent is a request, answered on someone else's timeline, in whatever shape they have decided to offer, and possibly at a cost, too. Our knowledge base is wired into our agents and in daily use. Connecting the CRM the same way is next, and is being built now.
And you can only put your own guardrails on your own systems. This is the one I wouldn't have predicted. Our knowledge base has a required review step: a person approves each entry before it can be used, and some entries get marked as incorrect, unnecessary, or confidential. Our agents only see what a person has approved, and what's marked confidential stays with the people cleared to see it. So when I'm building a proposal and I ask what we know about something, what comes back has been cleared by a person. The risk everybody rightly worries about with AI in a commercial function is a confident sentence about a capability you don't have being presented externally. The fix for that is a permission boundary, and you can only build one of those into a system you control.
That third point is really the whole argument. The systems holding what a company knows are exactly the systems its agents will lean on hardest, which makes them the ones most worth owning.
What we built, and what it changed
Our new knowledge base is where things started, and it is the one I would steal if I were reading this. We are a remote team and have been either remote or hybrid for years. We were getting great information about our latest development explorations and points of view from our colleagues on calls, but those were decentralized, hard to search, and difficult to use as a team responsible for creating content and putting together approaches.
My colleague, Rand, built a proof of concept that was hugely impactful. It quickly rose to the top of our priority list when Britton, now our Director of Agentic Operations, took it on. Britton and Rand built an agentic intake pipeline that ingests call transcripts our team uploads, reviews them and breaks them into specific content blocks, and triages them for manual human review before they can be referenced. The database is searchable, correctable, and now connected to our agents.
What it holds is what we have actually learned as a company: positions we have taken, things we have tried, how they went. Entries are approved by a person before they are logged, and some are marked confidential.
Our new CRM now does exactly what we need it to, because it is shaped to how we actually work. It has modules we've built to assist with how we do staffing planning, how we run campaigns, what's necessary for our reporting, and more. We're working on a connection now that our agents can use, to make updating records easier and to capture rich details that might have been glossed over if done manually.
Two big things changed for the better:
The first is about a remote team. We are distributed, and what we knew used to live in the heads of whoever had been in the room. A colleague's hard-won read on a market, or team discussions about new and different approaches, was recoverable only by finding that person and asking. Now it's written down, approved, and there for anyone on the team to find when they need it, rather than when they remember to ask. I'm a better leader and collaborator because I have a deeper understanding of our work, our wins, our losses, and the "whys" behind decisions made over multiple years as a team, because the material we fed it goes back further than the system does.
The second is about external-facing work. The materials we create (blog posts, proposals, ad campaigns) can now cite something verified rather than something remembered. It also means the work can move between us. If someone's away or stretched thin, whoever picks it up isn't starting from scratch.
Why this might matter to you
Something has changed about the value of what a company knows. Every company can now reach roughly the same models. What differs is what you point them at: the reasoning behind a decision three years ago, the thing your best salesperson understands and has never written down, trends in the impact we've driven for our clients over 13 years of work. That is the only input your competitors don't have.
Which means the systems holding what you know have become the most strategic software you own. And for most companies that information sits in someone else's product, in a shape somebody else designed, in a schema you can't extend, and reachable by your own employees and agents only in whatever way that vendor has decided to offer.
The second thing that has changed is the value of being able to alter your own systems quickly. That used to be a convenience. When your team is learning how to work with agents, and the requirements genuinely change month to month, the speed at which you can reshape your own tooling becomes a ceiling on how fast your team can get good.
We did not do this to save money, and I wouldn't recommend it to anyone whose main problem is their software budget. The honest reason to look at it is that the complexity of a stitched-together stack compounds quietly, and the day you want to make a real change is the day you find out how much of your own operation you can't actually alter.
Where we actually are
I want to be precise about the timeline, because a year sounds like a long time. We have been building toward this for about a year. We have only been using both systems regularly for a couple of months.
Both are in daily use now, not in pilot. The CRM is used across the organization every day, and connecting it to our agents is in the works. The knowledge base is used daily too, and our agents are already connected to it.
One honest caveat: the software, the data model and the workflows in our CRM are ours, but its database still runs on a hosted service – we're partway through moving it onto our own platform, and our knowledge base has already made that move.
And how much each of us use these new tools and workflows varies. Some of us are further along than others, and the good ideas one person has found aren't yet shared consistently across the board. So we are working on a playbook and a set of agentic directions to connect teammates' good ideas to one another, codify what has worked, and keep improving how we work together remotely.
That is the honest state of it: the tooling is now ahead of the habits. What I notice most is that the tooling now meets us where we are as human users and how our organization actually works, rather than the other way round. The agentic habits and workflows are the part we are still building and will continue to evolve.
If you are weighing something similar, the advice I'd give is specific. Don't start by asking what you could build. Start by asking which one or two of your systems hold information that your team could use more effectively in order to ship better work. The ability to own, change, connect, and use that information is what lets your team move as fast as it's learning.
If you are thinking about building your own tools like this and want to compare notes, please reach out!
I am Adrienne Marshall, head of sales and marketing at Rangle. Email me directly at adrienne.marshall@rangle.io.
