TL;DR
The 47 checks split across six categories:
- Data foundation (8 checks): Do you have clean data AI can work with?
- Workflow documentation (7 checks): Are your processes documented enough to automate?
- Tool stack (8 checks): Does your existing stack play nicely with AI?
- Team capability (8 checks): Can your team actually adopt what gets built?
- Volume threshold (6 checks): Do you do enough of the thing to justify automating?
- Commercial fit (10 checks): Will AI actually move your revenue needle?
Score 35+ out of 47 = ready. 25-34 = needs prep work first. Under 25 = AI is not your bottleneck.
Category 1: Data foundation (8 checks)
This is the most boring category and the most important. AI without clean data is a hallucination machine.
Check 1: Do you have a single source of truth for customer data?
CRM, spreadsheet, even a Notion database. One place where the canonical version of a customer record lives. If you have customer data scattered across Mailchimp, HubSpot, a Google Sheet, and three salespeople's heads, AI will multiply the chaos, not resolve it.
Pass criteria: One system holds the canonical record. Everything else syncs to or from it.
Check 2: Is your email list deduplicated and segmented?
Run a quick check. How many duplicate email addresses do you have? How many segments exist beyond "everyone"? If the answers are "lots" and "none," you're not ready to use AI for email personalisation.
Pass criteria: Under 5% duplicates. At least three meaningful segments.
Check 3: Do you have at least 12 months of conversion data?
AI models need patterns to learn from. Three months of conversion data is too thin. Twelve months captures seasonality.
Pass criteria: 12+ months of conversion data with date stamps and source attribution.
Check 4: Is your website analytics actually installed correctly?
Specific check: pull up your GA4 (or whatever you use) and verify that goals are firing correctly. Most analytics setups I audit are broken in at least one place. Goals firing for navigation events. Duplicate conversion tracking. Bot traffic counted as conversions.
Pass criteria: Conversion tracking verified working, bot filters in place, attribution model documented.
Check 5: Do you have product or service taxonomy documented?
For e-commerce: clean product categories with consistent attributes. For services: clear service definitions with delivery scope, pricing tiers, exclusions. If your services are "whatever the client needs," AI cannot help you scope them.
Pass criteria: Written taxonomy. Categories don't overlap. Attributes are consistent.
Check 6: Are your content assets indexed and searchable?
Past blog posts, sales decks, case studies, email copy, FAQs. If they're scattered across Dropbox, Google Drive, ten people's laptops, and an old WordPress install, AI cannot RAG them effectively.
Pass criteria: Central content repository with consistent file naming and metadata.
Check 7: Do you have unstructured customer feedback available?
Sales call transcripts. Support tickets. Cancellation reasons. Social mentions. AI is genuinely brilliant at synthesising unstructured customer voice if you have it. If you don't, you're missing the most valuable input.
Pass criteria: At least 100 pieces of unstructured customer feedback available in text form.
Check 8: Is anyone responsible for data quality on an ongoing basis?
Not "we'll get to it." A named person, a recurring meeting, a documented process. AI will degrade in production if data quality degrades. Someone needs to own it.
Pass criteria: Named owner. Defined cadence. Documented remediation process.
Category 2: Workflow documentation (7 checks)
You cannot automate what you cannot describe. This category catches the businesses that THINK they have processes but actually have habits.
Check 9: Is your lead-to-customer journey documented end-to-end?
Every step from inbound enquiry to closed deal to onboarding to delivery. Written down. With owners and SLAs. Most businesses I audit have this in three people's heads, written down nowhere.
Pass criteria: Written documentation. Each step has an owner. Each handoff has an SLA.
Check 10: Are your content workflows documented?
Who briefs. Who writes. Who edits. Who publishes. What the brief format is. What the editorial standards are. If this is "Sarah does it," AI can't slot in.
Pass criteria: Written workflow with roles, handoffs, and standards.
Check 11: Do you have prompt libraries or template responses?
For sales outreach, support replies, social media, content briefs. Even unsophisticated templates count. If every response is written from scratch, you're paying a tax that AI could remove tomorrow.
Pass criteria: At least one documented template per recurring response type.
Check 12: Is decision authority documented?
Pass criteria: Written approval matrix. Updated within the last 12 months.
Check 13: Are your reporting cadences fixed?
Weekly dashboards. Monthly reviews. Quarterly strategy. If reporting is ad hoc, you cannot measure whether AI improvements moved the needle.
Pass criteria: Documented reporting calendar. Reports actually happen on schedule.
Check 14: Do you have a feedback loop from delivery back to marketing?
What clients actually buy, why they churn, what they upsell into. If this never gets back to the people writing the marketing, your marketing optimises for the wrong things and AI accelerates the wrong direction.
Pass criteria: Documented mechanism for delivery insights to reach marketing. Happens at least quarterly.
Check 15: Is your meeting cadence sane?
If your team is in meetings 30+ hours per week, AI cannot help. Adoption requires time to implement. Time requires meetings get cut. If meetings can't be cut, you're not ready.
Pass criteria: Average team member has 15+ hours/week of focused work time.
Category 3: Tool stack (8 checks)
Your existing stack determines what's easy and what's expensive when adding AI.
Check 16: Does your CRM have an API or Zapier-equivalent integration?
If you're on a CRM that doesn't expose data programmatically (some legacy systems still don't), AI integration costs 5-10x more. Salesforce, HubSpot, Pipedrive, Close — all fine. Anything home-grown without an API is a red flag.
Pass criteria: CRM has API access OR is in a major integration platform.
Check 17: Is your email platform AI-friendly?
ConvertKit, Klaviyo, ActiveCampaign, Mailchimp — fine. Some older platforms have terrible exports and no API. If you can't get your email data out in CSV or via API, you can't run AI on it.
Pass criteria: Email platform allows CSV exports AND API access.
Check 18: Do you have content management you control?
WordPress, Webflow, Ghost, even Squarespace — fine, you control the content. If your site is hosted on a closed platform without admin access to the database or post objects, AI content workflows are limited.
Pass criteria: You can programmatically push and edit content.
Check 19: Is your file storage centralised and AI-accessible?
Google Drive, Dropbox, OneDrive — all integrate. If your files are on local drives or a closed corporate share, AI cannot reach them.
Pass criteria: Files in a cloud storage system with API/permissions access.
Check 20: Do you use a communication tool with thread persistence?
Slack, Microsoft Teams, Discord. AI works much better with persistent threads it can search and learn from. Email-only communication is harder to mine.
Pass criteria: Team uses a persistent-thread communication tool for work conversations.
Check 21: Is your project management system actually used?
If you have Asana/ClickUp/Monday but nobody updates it, AI cannot reason about your work. If your work happens in email and Slack DMs, it's invisible to AI.
Pass criteria: PM system reflects reality. Updated within the last 7 days for any active project.
Check 22: Is your data warehouse situation honest?
Most small businesses don't have a data warehouse. That's fine. But if you THINK you do and it's actually a Google Sheet that broke last quarter, fix that before adding AI.
Pass criteria: Either a real warehouse (BigQuery, Snowflake) OR honest acknowledgement you don't have one.
Check 23: Are you running AI tools nobody knows are running?
I audit this constantly. Three different ChatGPT subscriptions. Four people using different writing tools. Two competing automations. Before adding more AI, inventory what's already running and consolidate.
Pass criteria: Documented inventory of every AI tool in use. Single owner per tool.
Category 4: Team capability (8 checks)
AI is adopted by humans or it isn't adopted at all.
Check 24: Does your team have at least one person who's comfortable with prompts?
Not an engineer. Someone who's spent time writing prompts and iterating on outputs. If nobody on the team has this skill, you'll need to hire or develop it before adoption sticks.
Pass criteria: At least one team member can demonstrate a prompt that consistently produces useful output.
Check 25: Does the team have time to learn?
Adopting AI tooling costs 20-40 hours per person upfront. If everyone's at 110% capacity, you'll buy tools nobody uses.
Pass criteria: At least 4 hours per week per relevant team member earmarked for adoption.
Check 26: Is there a senior sponsor?
AI rollouts fail at scale when nobody senior owns them. Not "the CEO is excited about AI." A named senior who's measured on adoption and outcomes.
Pass criteria: Named senior sponsor. AI adoption appears in their objectives.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Check 27: Has your team experienced any AI tool successfully?
If the team has only tried AI for fun, adoption is harder. If they've already had a small win (transcription, summarisation, anything), they know it works.
Pass criteria: Documented small win with AI tooling in the last 6 months.
Check 28: Are you willing to deprecate work that AI replaces?
Honest question. If AI removes the need for a manual report someone takes pride in, are you willing to delete that report? If not, AI gets layered on top of existing work and the team burns out.
Pass criteria: Written agreement that AI-replaced work will be deprecated, not preserved.
Check 29: Is the team rewarded for adoption?
Not punishment for non-adoption. Reward for using the tools. If using AI feels risky and not using AI feels safe, nobody uses AI.
Pass criteria: Documented incentive (recognition, time saved counted toward goals, etc.).
Check 30: Are you prepared for some attrition?
AI rollouts surface team members who don't want to adapt. Some leave. If you can't tolerate that, slow the rollout.
Pass criteria: Honest acknowledgement that some team turnover is acceptable.
Check 31: Is there someone responsible for AI ethics and quality?
Not a committee. A person who reviews what's being shipped and catches the embarrassing outputs before they go public. Without this role, you ship an AI-generated comment that mentions a competitor or hallucinates a stat.
Pass criteria: Named reviewer. Documented review process before public outputs.
Category 5: Volume threshold (6 checks)
AI economics depend on doing things repeatedly. Low volume = no payback.
Check 32: Do you send 1,000+ marketing emails per month?
Below 1,000, the manual approach is fine. AI personalisation pays back at scale.
Pass criteria: 1,000+ monthly email sends.
Check 33: Do you publish 4+ pieces of content per month?
Below 4, an AI content workflow is overkill. At 4+, the workflow saves real hours.
Pass criteria: 4+ pieces of substantial content per month.
Check 34: Do you handle 50+ inbound leads per month?
Below 50, manual handling works. Above 50, AI qualification and routing earns its keep.
Pass criteria: 50+ monthly inbound leads.
Check 35: Do you have 100+ recurring customer interactions per month?
Support tickets, account check-ins, success calls. Below 100, automation isn't worth it. Above, it is.
Pass criteria: 100+ recurring customer interactions per month.
Check 36: Do you have 10+ sales calls per week?
Sales call transcription, summary, and follow-up automation pays back above this threshold.
Pass criteria: 10+ weekly sales calls.
Check 37: Is the volume growing?
If you're at the threshold but flat, AI is borderline worth it. If you're growing, you'll cross every threshold inside 6-12 months and the prep work is worth doing now.
Pass criteria: Volume growth of at least 20% year-over-year on your primary metric.
Category 6: Commercial fit (10 checks)
Even with everything else in place, AI might not move your specific revenue needle.
Check 38: Is your primary bottleneck addressable by AI?
If your bottleneck is "we can't get on sales calls because the team is too small," AI helps. If your bottleneck is "the founder won't delegate," AI doesn't help. Diagnose the bottleneck honestly first.
Pass criteria: Documented bottleneck. AI demonstrably addresses it.
Check 39: Do you have margin to invest?
AI projects pay back in 6-18 months. If you're cash-strapped and need ROI in 60 days, AI is the wrong investment.
Pass criteria: 6+ months of runway available for the AI investment without compromising operations.
Check 40: Is your average customer value high enough?
Pass criteria: Documented AOV. Realistic uplift calculation that pays back inside 18 months.
Check 41: Are competitors already moving on AI?
If your competitors aren't moving, you have time to be deliberate. If they are, you have less. This affects pacing, not whether you do it.
Pass criteria: Documented competitive intelligence on AI adoption in your space.
Check 42: Do you have customers who care?
For some segments, AI-augmented service is a buying signal. For others, it's a turn-off. Check with actual customers before assuming.
Pass criteria: At least 5 customer conversations explicitly covering their view on AI in your category.
Check 43: Is your pricing model AI-compatible?
If you bill hourly for delivery work and AI cuts the hours by 70%, your revenue drops. If you bill on value or outcome, AI accelerates margin. Fix the pricing model before AI eats your business.
Pass criteria: Pricing model where AI efficiency increases (not decreases) revenue per customer.
Check 44: Have you costed the build properly?
Not the tool subscription. The implementation cost. The training cost. The maintenance cost. The integration cost. Most AI ROI calculations only count the subscription and overstate ROI by 4-5x.
Pass criteria: Documented total cost of ownership including all categories above.
Check 45: Do you have a measurement plan?
How will you know AI worked? What's the baseline? What's the target? When do you re-evaluate? If you can't answer these, you'll spend money and never know if it paid back.
Pass criteria: Documented measurement plan with baseline, target, and review cadence.
Check 46: Have you stress-tested the failure case?
Pass criteria: Documented worst-case scenario. Business survives it without crisis.
Check 47: Is the AI work explicitly on the strategic roadmap?
If AI is a side project somebody snuck in, it gets de-prioritised the first time something more urgent happens. If it's on the roadmap with a name next to it, it survives.
Pass criteria: AI initiatives appear in the documented strategic plan with named owners and milestones.
How to use this audit
Score each check pass/fail. Tally the passes.
35-47: You're ready. Start with the highest-volume, highest-friction workflow first.
25-34: Spend a quarter on prep work. The biggest ROI from your AI budget will come from fixing whatever you scored below 5 on. Document workflows, clean data, fix pricing, before you spend on AI tools.
Under 25: AI is not your bottleneck. Other interventions (hiring, process, positioning) will pay back faster. Come back to this audit in 6-12 months.
I run this audit on every business before I'll quote them for implementation. About 30% pass first time. About 40% need a prep quarter. About 30% I send away because AI genuinely won't help them yet, and saying so costs me the engagement but saves them the cheque.
Frequently asked questions
How long does a full audit take? About a day to run yourself if you have the documentation. 90 minutes if I run it with you, because I know what to ask. The real time goes into fixing what fails, not running the audit.
Do I need an external auditor? No. The framework above is everything. The reason to bring in an outside view is honesty — you'll grade yourself more generously than the framework intends.
What's the most common failure point? Workflow documentation. Most businesses have habits, not processes. AI exposes this immediately.
Can I skip checks that don't apply? No. If a check doesn't apply, score it as pass. But verify it genuinely doesn't apply, don't skip because it's inconvenient.
How often should I re-run the audit? Annually. Or whenever you're about to make a major AI investment.
Want me to run this with you?
If the result is "wait," you walk away with a fix list and no further pitch. That's the deal.
I'm Lilach Bullock. I've been a marketing consultant for twenty-one years. I went all in on AI in 2024. I work with founders and marketing leaders who want AI to actually change their numbers, not just their tool stack.
Related from this week
- /ai-implementation-roadmap/ — the 90-day path I actually use
- /ai-for-marketing-agencies/ — AI for agency operators specifically
- /ai-for-seo-without-penalty/ — the AI-safe SEO playbook
Related: AI Prompts for Marketers, The 30 I really Use (2026)
