Today I gave my first session on intelligence layer design with my development team at Kolaborate. In that session, I took them through the concept of touch points — an idea I've been developing that I believe is fundamental to how we should think about data, automation, and intelligent systems.
I want to elaborate further on this idea by fully delving into what touch points are, and how they connect to what I'm calling synaptic data-driven workflows.
What Is a Touch Point?
A touch point, in simple terms, is a point in time at which data is manipulated. That's it.
Why is this important? Because how and when data is manipulated is a tiny snowball that leads to an avalanche within any database.
Take, for example, the manipulation of a single figure in a financial database. Consider how that one change reflects across all reports, all calculations, and everything related to that data point within the broader context of an intelligence layer. One small mutation can cascade through an entire system.
This concept becomes even more powerful when you consider that we are now capable of having AI watch for changes in data and take intelligent, action-based responses around the manipulation that has occurred — all while factoring in the precise time at which that manipulation took place.
Touch points open doorways to a whole new range of potential outcomes. Some of these outcomes could be intrinsic to the intelligence layer itself — meaning the output of an intelligent action, based on the data that was manipulated, is then fed back to further refine the agent performing that action. This creates a feedback loop that we could think of as a form of metacognition: the system reasoning about and improving its own reasoning.
Chaining Touch Points: Synaptic Data-Driven Workflows
Now things get interesting when you start chaining touch points together. When you do this, you form what I would like to call data-driven synaptic workflows.
I know this terminology might sound bold, but bear with me. We understand what a synapse is by definition — a junction where signals pass between neurons in a network. I think this is a powerful and accurate term to describe what is taking place when you chain together intelligent touch points.
When I say intelligent touch points, I am not referring to artificial intelligence specifically. I am talking about a genuinely intelligent touch point — one that involves human-level intelligence in its design and implementation. The intelligence is in the architecture, the forethought, the deliberate construction of what should happen when data changes.
A Concrete Example
Let me illustrate this with a scenario. Imagine you have four different data points, all within a financial database. You already have basic touch points in place — for instance, if your profit margins drop below a specific threshold in the database, then a predefined intelligent action is triggered.
That's useful, but it's a single touch point with a single response. Now imagine upgrading that simple implementation into a more robust synaptic workflow:
When our profit margins reach a threshold that warrants a warning, I want the system to call our Head of Finance on his direct phone number and read out to him the current state of our finances — explaining specifically why it is urgent that we make a particular decision. Once that call is completed, I want the system to notify the CEO and CTO that this communication has taken place, and attach to that notification verifiable proof of why these actions were triggered.
That is a chain of intelligent touch points: data change triggers analysis, analysis triggers direct human communication, completion of that communication triggers notification with an auditable evidence trail. Each link in the chain is a synapse — a deliberate, designed junction where one intelligent action passes its signal to the next.
The Future: Intelligence Layer as a Service
With the emergence of agentic AI and agentic workflows, intelligence layers are now a reality. I strongly believe that we are heading towards a fundamental shift in what the software industry offers. We will move from traditional software-as-a-service to what I would describe as Intelligence Layer as a Service.
Here is why: with the right data-driven synaptic workflows, you could design an intelligence layer that manages its entire infrastructure — websites, domains, emails, databases, and everything in between — on its own, and progressively gets better at doing so.
At that point, you no longer need humans to build the underlying software layer. Instead, you need humans to design, maintain, and refine the intelligence layer that is operating your business on your behalf. The role of the engineer shifts from building software to orchestrating intelligence.
This is my first deep dive into what I think is going to be one of many explorations on this topic. I have a lot more to share, and I'm looking forward to developing these ideas further as we continue building intelligence layers in practice at Kolaborate.