The context AI needs. Kept by the firm, not the vendor.
AI can design, troubleshoot and operate networks, and at many tasks it is very good. What it cannot do is know why your estate is the way it is. CIK keeps that knowledge in one place that your engineers and your AI tools both read, checks it against what is running, and ties every piece of it to the business outcome it serves.
AI is already useful in network engineering. It drafts designs, reads logs faster than any person and explains configuration it has never seen. This site does not argue with any of that.
AI is only as good as what it knows about your estate.
Why it was built that way, what was traded off, and what must not change. Most of that knowledge sits in engineers' heads, and some of it has already left with them.
Without it, even the best model gives confident answers that are right in general and wrong for your firm.
The same 20 questions, with and without the firm's records.
Why a setting is the way it is, what a change would break, who owns an exception. Same model, same questions; the only difference is the records.
“Yes, adding VNI 70 without a VRF is consistent with the current design.”
“The repository does not name anyone.”
DeepSeek V4.1 Flash, first run; every answer read against the records. 20 questions, written by the people who built the records. See the full comparison
Capture, connect, check.
-
STEP 1
Capture
The reasoning of the engineers who were here, who are here and who come next: each decision, what it was weighed against, who made it and what it assumes.
-
STEP 2
Connect
In a structured form. Each decision links up to the business outcome it serves and down to the configuration that implements it. People and AI tools both read from it and add to it.
-
STEP 3
Check
Continuously, against what is running. When the network, the records or the business requirement drift apart, the difference is flagged.
Then back to capture: every change and every run goes round the loop again.
Three things you can hold the work to.
One source for people and AI
The engineer and the agent get the same answer, from the same records.
Always checked
Drift between the network, the records and the business requirement is flagged, not discovered in an outage.
It stays with the firm
The knowledge lives in the firm's own repositories, in an open format any AI tool can read. Changing AI vendor does not mean losing what the firm knows.
Automation and AI are not the goal. Every piece of work answers one question: for which outcome?
Each outcome has an owner, a metric, where it stands today and a dated target.
Work that serves no outcome is flagged as cost.
An outcome with no work under it is flagged as an unfunded promise.
Most firms do not start this because it is a good idea. They start because something happened.
- An audit finding Evidence produced in a scramble, findings that recur Control
- An incident with a manual-error cause Outage minutes, trust in the change process Resilience
- A merger or consolidation Two estates, one team, no shared memory of either Scale
- A cost mandate Spend that nobody can tie to an outcome Cost
- A key person leaving Decisions only one person could explain Continuity
Start with one design.
Get in touch