Know What's Actually Possible
Get realistic timelines for common marketing AI projects. Learn about hidden delays, what AI can and can't deliver, and how to set expectations that stick.
The gap between what vendors promise and what's technically feasible is where budgets go to die. Here's a reality check on timelines, capabilities, and hidden delays.
REALISTIC TIMELINES FOR COMMON REQUESTS
Simple API Integration (pre-built connector): 2–5 days. Why: Testing and edge cases take time even with "plug-and-play" solutions. Red flag: "Instant setup" but your system needs custom fields.
Custom API Integration: 3–6 weeks. Why: Development, testing, and security review. Red flag: Vendor says "2 weeks" without asking about your security requirements.
AI Chatbot (using a template): 1–2 weeks. Why: Data upload, knowledge base organization, initial testing. Red flag: Claims "48 hours" but skips the training and testing phase.
AI Chatbot (custom-built): 6–12 weeks. Why: Custom training, knowledge base creation, conversation flow design, edge case handling. Red flag: Vendor doesn't ask about your content sources.
Data Migration (CRM, MAP, or CDP): 4–12 weeks. Why: Field mapping, data cleaning, validation, deduplication. Red flag: Vendor doesn't ask about your data quality.
Landing Page Personalization (rules-based): 1–2 weeks. Why: Segment definition, rule setup, QA. Red flag: Claims "real-time AI" but it's actually if/then logic.
Landing Page Personalization (AI-driven): 6–8 weeks. Why: Model training, learning phase, A/B testing. Red flag: Doesn't mention a "learning phase" or minimum traffic requirements.
Email Automation (simple triggers): 3–5 days. Why: Trigger configuration, template creation, testing. Red flag: No mention of deliverability testing.
Email Automation (complex journeys): 2–4 weeks. Why: Logic mapping, branching paths, edge case handling. Red flag: Doesn't ask about your current workflows.
HIDDEN DELAYS THAT VENDORS DON'T MENTION
1. Your IT Security Review (add 2–4 weeks) Any new tool touching customer data requires security approval. The vendor doesn't know your internal process. Ask your team: Does this need a security review? How long does our approval process take?
2. Data Quality Issues (add 1–8 weeks) AI and machine learning fail with dirty data. Duplicates, missing fields, and inconsistent formatting all need cleaning. Ask: What's the current state of our CRM data? Do we have a data cleanliness score?
3. Stakeholder Approval Workflows (add 1–4 weeks) Most B2B purchases need 3–5 stakeholders: IT, Security, Legal, Finance, and an executive sponsor. Vendor timelines assume instant internal approval. Ask: Who needs to sign off? What's a realistic timeline for getting everyone aligned?
4. Change Management and Training (add 2–6 weeks) Your team needs to learn the new tool. Vendor timelines rarely include training. Ask: What training is included? How long until our team is self-sufficient?
CAN WE ACTUALLY BUILD THIS?
Things AI can do (with realistic caveats): Write a weekly newsletter: Yes — but draft-only. Requires human editing. Expect 30–40% time savings, not 100%. Auto-personalize the website in real-time: Yes — but needs a CDP and personalization engine. Cost: $50K–$200K per year. Timeline: 3–6 months to see ROI. Predict which leads will convert: Yes — if you have 1,000+ historical data points. Needs quarterly retraining. Timeline: 2–3 months to build and validate. Integrate CRM with marketing platform: Yes — via APIs or tools like Zapier. Cost: $0–$25K. Timeline: 2 days (Zapier) to 6 weeks (custom).
Things AI can't do (despite vendor claims): Create social content AND auto-post with no review: AI can draft, but no AI truly understands trending context or brand nuance in real time. Use AI for drafts, humans for approval and posting. Build a chatbot that answers ANY question perfectly: Accuracy typically sits at 70–85%. Requires a comprehensive, well-maintained knowledge base. Always needs a human fallback option. Timeline: 4–8 weeks just for training.
WHAT YOU'VE LEARNED ACROSS ALL 5 LESSONS
You now have a complete framework for evaluating AI marketing technology:
S — Signal from Noise: The 3-Question Filter to separate real AI from rebranded software. T — Translate Claims: 12 technical terms decoded so vendors can't hide behind jargon. A — Assess AI: The Capability Ladder to evaluate what AI actually does well versus what it can't. C — Challenge Vendors: 7 red flags and a demo checklist to protect your budget. K — Know What's Possible: Realistic timelines and honest capability assessments.
You're now equipped to evaluate martech vendors without getting lost in jargon, challenge unrealistic timelines, spot vendor theater, and make smarter budget decisions with confidence.