TL;DR
A martech AI readiness audit measures four layers: data, integration, AI features already paid for, and governance. Most brands are already paying for 70% of the AI capability they need; they just have not turned it on. The audit protocol is catalog, classify, kill, activate, scored on a 0-5 rubric per layer, with a one-pager as the output. Skip vendor demos. The work is inside the stack you already own.
- Four layers, scored 0 to 5, independently.
- The 70% rule: most AI you need is already paid for.
- Catalog, classify, kill, activate, in that order.
- The output is a one-pager, not a deck.
- Re-run the audit yearly. The stack drifts.
In this article
The four-layer audit
A martech AI readiness audit is not a vendor inventory. It is a four-layer assessment of whether your stack can actually run AI in production. The four layers are independent. A high score on one does not compensate for a zero on another.
Layer 1: Data
Data is the foundation. The questions to answer: do you have a single source of truth for customer data, do you have first-party event data piping into a place you can query, are the schemas documented, is the data clean enough to be used in a model prompt without further engineering?
If the answer to the last one is no, every AI initiative downstream is on a clock. AI multiplies whatever your data quality is. Bad data plus AI equals confident, fast, articulate wrong answers, at scale.
Layer 2: Integration
Integration is whether the systems talk to each other in real time, or close to it. Can your CRM, ESP, ad platforms, support tool, and product analytics all be queried from the same workflow? Are the integrations API-first or are they stuck in nightly CSV land?
Integration debt is the silent tax on every AI project. The work that should take a week takes a month because the data is in four systems and none of them have an API your team can use.
Layer 3: AI features already paid for
This is the layer most teams skip and most savings sit in. Every major martech vendor has shipped AI features into their existing products: predictive sends, generative copy, segment recommendations, lookalike modeling, anomaly detection in analytics. Most brands have these features sitting dark on tools they already pay for.
Cataloging the AI features you are already paying for is usually the highest-ROI hour of the entire audit.
Layer 4: Governance
Governance is whether you can actually deploy AI without your legal, privacy, security, or brand teams blocking the launch. The questions: do you have a vendor AI policy, do you have approved data residency, do you have a brand voice spec, do you have a content review workflow that does not bottleneck on one person?
Governance is usually the layer that sounds boring and turns out to be the gating constraint. Programs stall here more often than on capability or data.
The 70% rule
Across the audits I have run at Automatic and across portfolio diligence in past roles, a consistent pattern holds: most consumer brands are already paying for roughly 70% of the AI capability they actually need. They just have not turned it on.
The 70% breaks down predictably. The ESP has predictive send-time, AI-driven segmentation, and a generative copy assistant. The CDP has lookalike modeling and predictive churn. The analytics tool has anomaly detection. The CRM has generative summaries. The ad platforms have bid automation and creative AI. Most of these features sit dormant.
Why do they sit dormant? Three reasons:
- The team that bought the tool is not the team using the AI feature. The AI feature shipped after procurement, in a release note nobody read.
- The features are not on by default. They require activation, configuration, and someone to own the output.
- The team is being marketed to by net-new AI vendors, who promise the same capability the existing tool already offers, more loudly.
The first move after the audit is to activate what you own, not to buy what you do not.
Most consumer brands do not have an AI tooling problem. They have an AI activation problem on tools they already paid for.
The audit protocol: catalog, classify, kill, activate
The audit runs in four stages over two to four weeks. Each stage produces an artifact.
Stage 1: Catalog (week 1)
List every martech tool your company pays for, with: annual cost, owner team, primary use case, contract end date, and a yes/no on whether it has AI features shipped or on the roadmap. Pull invoices, finance records, and IT records. The list will be longer than the team thinks. The deltas between the lists are the real finding.
Stage 2: Classify (week 2)
Classify each tool on three dimensions:
- Strategic role: core, supporting, or redundant.
- AI capability: AI-native, AI-augmented, or AI-absent.
- Activation status: activated, dormant, or unknown.
Score each tool on the four layers above. Roll up to a stack-level readiness number per layer.
Stage 3: Kill (week 3)
Build the kill list. Tools that are redundant, dormant on AI features you needed, or unused by the owning team go on the list. Tools whose contract end date is within two quarters get extra scrutiny because the renewal decision is approaching anyway.
The kill list almost always frees enough budget to pay for the activation work in stage 4. That is the financial story for the leadership team.
Stage 4: Activate (week 4)
Build the activation list. Pick the top 5 to 10 AI features you are already paying for and queue them for activation, with a named owner per feature. Pair this with the readiness one-pager and present to leadership.
Scoring AI readiness on a 0-5 scale
Score each of the four layers independently on a 0-5 rubric. I use the same scale across every audit because the comparison across layers is the insight.
- 0: Missing or broken. The layer does not exist as a coherent capability.
- 1: Ad hoc. Something exists but it is undocumented and inconsistent.
- 2: Defined but inconsistent. There is a policy or system, but execution is uneven.
- 3: Reliable. The capability works most of the time and is owned by a team.
- 4: Operational. Documented, owned, measured, with operating cadence.
- 5: Production-grade. AI workflows can run on this layer without engineering work each time.
Most consumer brands score themselves at 2-3 across the board on a first audit. That is normal. The point of the score is not to feel bad about it. It is to know which layer to fix first. Usually the lowest layer is the gating constraint, and fixing it lifts everything above it.
The readiness gaps that matter (and the ones that do not)
Some gaps stop the program. Some gaps are theater. Telling them apart is the audit's actual value.
Gaps that matter:
- No single source of truth for customer data. Everything else falls apart downstream.
- No real-time or near-real-time integration between the systems your AI workflows need to touch.
- No vendor AI policy. Legal will pump the brakes when you try to ship.
- No named owner of the marketing data layer. Without one, nothing gets fixed.
- No brand voice spec written down. Every AI content output is going to drift.
Gaps that do not matter:
- "We do not have a CDP yet." Many brands run fine without one if their data warehouse is in shape.
- "We do not have a dedicated AI tool." See the 70% rule.
- "We have not picked an LLM." Pick last, not first. See how to pick the right LLM.
- "We do not have an AI committee." Committees are not capabilities.
The output: the readiness one-pager
The deliverable is a one-pager. Not a deck. One sheet of paper, designed to be read in two minutes by the CEO.
The one-pager has six sections:
- Overall readiness number, 0 to 5, computed from the four layer scores.
- The four layer scores, with one sentence each on the gap.
- The 70% line, showing the dollars of AI capability already paid for and dormant.
- The kill list, with annual savings and contract end dates.
- The activation list, with named owners and 90-day milestones.
- The next-quarter ask: the resources, decisions, and approvals needed.
The one-pager is the artifact that anchors the program for the next quarter. It feeds into the broader AI transformation playbook and the V1 Framework's "define done" discipline. If you cannot write the one-pager, the audit is not complete.
If the audit cannot be summarized on one page, the audit is the problem, not the stack.
The bottom line
Audit before you buy. Activate before you license. Score independently across the four layers. Build the one-pager. Then move.
The teams that run a real audit free budget, ship faster, and stop being marketed to by every AI vendor with a sales motion. The teams that skip the audit pay for tools they do not use and miss the AI features they already own. Pick the first one.
FAQ
What is martech AI readiness?
Martech AI readiness is the degree to which your marketing tech stack can actually run AI in production: clean data, real integration, AI features turned on inside the tools you already own, and governance that lets you ship. It is not a vendor checklist.
How long does a martech AI readiness audit take?
A focused audit runs in two to four weeks. Week one is catalog. Week two is classification and scoring. Week three is the kill list and the activation list. Week four is the readiness one-pager and the leadership readout. Anything longer is the audit becoming the project.
Who should run a martech AI readiness audit?
A single named owner with cross-functional access. That is usually a VP Marketing, a Fractional CTO, a Director of Marketing Operations, or an outside consultant. It is not a committee. The output is owned by one person.
What scoring system works for AI readiness?
A 0-5 scale per layer: 0 missing, 5 production-grade. Score data, integration, native AI activation, and governance independently. Composite the scores into a readiness number, but always report the layer scores so leadership sees where the gaps are.
What do you do with the audit result?
Three artifacts. A one-pager that shows the readiness scores. A kill list of tools to cut. An activation list of AI features you are already paying for and can turn on this quarter. Everything else is noise.
How often should you re-audit?
Once a year minimum, or after any major martech change. The stack drifts faster than people realize. Vendors ship AI features quarterly. New tools sneak in through team-level purchases. An annual re-audit catches the drift before it becomes a year of waste.