CCT — AI Adoption
2026-05-16

AI adoption is progressing — progress, voices, and what's next.

Where teams stand today on hands-on AI use, in their own words. Real outcomes, the operating patterns we've found unlock the most value, and the playbook we're using to onboard the next set of teams and roles.

0
Teams engaged
0
Practitioners with documented results
0
Typical time-to-impact
0
Unlock principles codified
0
Step onboarding playbook
Data Platform · Analytics · Nova From skepticism to "best week in a while" Repeatable adoption playbook
Section 1

Two approaches to AI adoption

Two distinct strategies for bringing a person up to speed with AI — different shapes for cost, time-to-impact, and depth of learning. Both reference the same five-level AI Adoption Spectrum.

AI Adoption Spectrum

A five-level model of how teams progress with AI assistance — from passive code suggestions through to a fully-resourced workspace built around the AI-DLC methodology.

1

Auto Complete

  • Code suggestions
  • No process change
2

Vibe Coding

  • Conversational
  • No specs
3

Guided Dev

  • Steering files
  • Context engineering
4

Spec Driven

  • Reqs → Design → Tasks → Code
5

Workspace

  • Pre-built workspace
  • Role-specific skills + guardrails
  • AI-DLC methodology

The two approaches

Approach 1

Build the Skill

Walk people left-to-right through the spectrum: Vibe Coding → Guided Dev → Spec Driven → Workspace. Each level introduces new concepts, best practices, and techniques for improving quality and time-to-build.

What it builds
Durable foundational knowledge of how AI operates, prompt patterns, context engineering, and the trade-offs between speed and rigor at each level.
Cost shape
Investment is in the person. Slower time-to-productivity, but the learning is transferable across tools, repos, and roles. Higher risk that the person gets side-tracked chasing short-term tools and experiments along the way.
Approach 2

Build the Workspace

Drop the person into a workspace that's already set up with the context, skills, and guardrails for the work they actually do. The system carries the expertise; the person becomes productive on day one and absorbs the foundational knowledge over time.

What it builds
Immediate productivity. Foundational understanding accrues gradually through hands-on use.
Cost shape
Investment is in the environment. Significant upfront effort by an experienced builder to assemble the right context, the right skills, the right review checkpoints, and the right guardrails so the person can be as effective as possible without needing deep AI literacy first. Higher risk that the person makes mistakes they don't fully understand and needs help getting unstuck.
Complementary, not exclusive. A practical rollout combines both — build the workspace for the broadest audience to unlock quick wins, while putting a subset of practitioners (the ones who will own and evolve the workspaces) through the skill-building path. The unlock principles and the onboarding playbook later in this doc apply to both approaches.
Section 2

AI Coaching — progress by team

Voices and outcomes from people doing the work. Tabs below switch between teams.

A
Alex
Data Platform · One week of hands-on use
Shipping

Alex is using AI to read our existing MS SQL vendor load procedures and translate them into Python code that reads from S3 parquet files into our MS SQL data warehouse databases. The goal is to replace the SQL Load procedures so a Custom RDS MS SQL Instance isn't needed for L&S.

Wow. It learns quick.
I guess I'll let Claude drive. It's a mindset change lol…
I find myself talking to it like a human being a lot.

Converted Oasis Table & Everi CDM SQL loads to Python with data validations, running in a local environment. After one week of use — production-class data plumbing translated, not just prototyped.

R
Red
Analytics · Implementation engineer
Compounding wins
I absolutely love it.
Biggest benefit is not having to deal with the git/check-in process myself. It was such a frustrating experience that's completely seamless now because I just make Claude do it for me.
Debug speed

Sag Chip data deletion

PlayerDayMeasure_Slot records were being arbitrarily deleted. Claude found the issue in about a minute.

~1 minute vs. ~1 hour solo
Cross-repo refactor

Player-metadata default value

Instead of relying on memory (or pinging Mark), had Claude scan every repo for places the ID was used and update them accordingly.

Multi-repo, one pass
Automation in flight

ADT Tier configuration

Three-table update driven from a customer email. Claude reads the email and translates it into the data updates. ~15–30 minutes per project — stacks across every implementation.

Per-project savings, then multiply
Friction removed

Git / check-in

Previously his biggest day-to-day frustration. Now seamless — Claude handles the mechanics.

E
Ethan
Analytics · Tooling & team enablement
Shipping

Repeatable workspace setup. Lots of progress over the AI Pilot Project time period — created a repeatable method for workspace setup that enables other IA TC's to work in it. Red and Lexi were the first beneficiaries. Foundational for the unlock principles in Section 3.

L
Lexi
Analytics · Newest IA team member
Onboarding

AI-assisted onboarding. Just joined IA — using Claude heavily to ramp by asking questions directly against the repo (code + documentation). Onboarding signal: new hires can self-serve context that previously required a tenured engineer.

B
Brandy
Project Nova · After one week with Claude Code + AI-DLC
Energized
It's incredible. Definitely makes me more excited. It's amazing what it can do with the right context.
Honestly, this has been the best week of work I've had in a while. It's been really fun so far.

VP Product, hands-on in the workflow. Working through the AI-DLC method end-to-end on Insight Cash — proof the methodology is approachable for non-engineering leaders. Cultural signal as much as a delivery signal.

What Brandy shipped in week one

From first PR (2026-05-11) through end of week one (2026-05-17) — authored on GitHub as CCTBrandy.

ticket-type-maintenance cash-count-category-maintenance deposit-category-type vault-management-shell /cct-prototype skill
Section 3

Key aspects for maximum unlock

The common patterns that allow for maximum potential and speed to productivity. These showed up consistently across every story in Section 2.

1

Shared workspace structure

Everyone working from the same folder + file + MCP layout, synced to the appropriate data sources. No team should have to invent it from scratch.

2

Skills for repeated work

Codify the flow and the specific tasks people repeat into reusable skills, so the team isn't re-explaining the same thing to Claude every session.

3

Just enough context, captured as you go

Leave a trail of .md files so Claude has the "memory" of what's been done — and patterns surface naturally for the next person to pick up.

4

Do everything through Claude

Resist the urge to step outside the session for a "quick" manual fix — even when you think you'd be faster, you're starving the system of context.

5

Build the mountain of context

Keep capturing the things that aren't documented or only live in someone's head. Tribal knowledge becomes a renewable asset, not a single point of failure.

6

Establish guardrails

Make sure the skills and structure we're building are the right ones, established correctly — review gates, naming conventions, and shared standards.

Section 4

Context organization for different teams & roles

The steps for creating the best workspace for onboarding a new team or a new role.

1

Identify the information sources required for work

Repos, websites, documentation, OneDrive — everywhere the work currently lives.

  • Determine what belongs in AI-Context.
  • Build steering guides for workflows, standards, references.
2

Create the local-workspace instructions

A Markdown or Notion page with the instructions for Claude to materialize a local workspace populated with the sources from step 1. Repeatable for every new team member.

3

Build skills for repeatable tasks & workflows

The work the team does often becomes a skill, not a fresh prompt every time. This is where the compounding starts.

4

Decide what artifacts to save as work executes

Prioritize artifacts that deliver one or more of:

  • Decision records — rationale, alternatives, risks, and benefits.
  • Education & onboarding — resource for the next person to ramp.
  • Claude's memory — context that improves future recommendations.
  • Consider whether the output also belongs in web documentation for Product.
5

Establish guardrails

Review gates, naming conventions, and shared standards — so what gets built fits the structure and stays consistent across teams.

Section 5

Potential next steps

Where to point investment over the coming quarter — converting individual wins into team- and org-level compounding gains.

Roll out the shared workspace

Step 1–2

Take Ethan's workspace pattern from Analytics and turn it into the org default. Same folders, same MCPs, same sync points for every new team.

Expand the skill library

Step 3

Codify the repeated tasks each team is already doing (e.g., Red's ADT-Tier email-to-data flow) into reusable skills. Publish a shared catalog.

Capture the tribal knowledge

Step 4

Schedule structured knowledge-extraction sessions for the things that only live in people's heads. Convert to .md context Claude can reuse forever.

Land the AI Pilot

In flight

Wrap the TC's pilot, publish results, and feed lessons into the playbook before the next wave of teams onboards.

Formalize guardrails

Step 5

Document the review gates, naming conventions, and quality bars that keep skills and context consistent as adoption scales.

Tell the story externally

Communications

Use Brandy's experience and Red's quantified wins as proof points in customer, recruiting, and investor conversations.

Headline takeaways

Time-to-impact~1 week
Teams engaged3
Practitioners with documented results6
Unlock principles codified6
Onboarding playbook steps5
Sentiment signal"Best week of work in a while"