#149 — The rise of AI strategists

Your CTO manages infrastructure, data architecture, and core technology systems, while an AI strategist brings strategic vision to integrate AI into business processes and identifies use cases that drive measurable business value. If your startup is under 20 people and still validating product-market fit, start by upskilling one versatile technical person who can prototype rapidly. Once you're scaling beyond early stage and have multiple departments, consider a dedicated AI strategist who can break down silos between data, content, and decisioning across teams.

Generative AI is reactive and creates content based on prompts—think copywriting, image generation, or document summarization. Agentic AI is proactive and takes autonomous actions to achieve multi-step objectives without constant human guidance. For example, generative AI drafts your customer emails, while agentic AI independently determines test coverage requirements, generates test cases, schedules execution, identifies failures, and escalates only when human intervention is needed. Your AI strategist determines when to deploy each type based on your specific business metrics.

Use a four-quadrant framework: cost savings (quantify downtime reduction—one manufacturer achieved 95% prediction accuracy for equipment failures, showing positive ROI in nine months), revenue generation (an eCommerce AI recommendation engine increased average order value through A/B testing), risk mitigation (improved compliance and audit accuracy), and strategic value (enhanced competitive capabilities). The key questions are: what new projects did AI enable you to complete, what value do those projects deliver, and how did AI move specific business metrics—not just how much faster tasks became.

Cursor (Anysphere) achieved $100 million ARR in just 21 months with only 20 people by building an AI coding assistant and leveraging organic growth through developer communities—zero marketing spend. They hit $200M ARR without any PPC ads or outbound campaigns. Another example: boohooMAN used AI personalization for SMS campaigns and saw 5x returns in the UK, with birthday campaigns delivering 25x ROI. The pattern: these startups focused AI on solving real, painful problems (fragmented development workflows, generic marketing) rather than generic efficiency gains.

Hire when you're transitioning from AI experimentation to operationalization across multiple departments. For startups under 10-15 people, start with 1-2 versatile generalists (an AI/ML engineer plus a data scientist) who report to a technical founder. Once you have distinct business units that need AI integration, product teams that require governance frameworks, or you're fielding client/customer requests for AI capabilities, bring in a dedicated strategist. Warning: don't overhire too soon—validate your AI direction with lean prototypes before building a full team.

Prioritize bridge-building over pure technical depth. Look for fundamental understanding of how models think and function (not just prompt engineering), ability to design agents and roadmaps that link departments, track record of measuring outcomes rather than just efficiency, and demonstrated curiosity through continuous learning. Communication skills matter enormously—they need to explain complex AI concepts in plain language and show empathy for employees anxious about automation. Check if they can articulate the difference between generative and agentic AI applications, understand governance and ethics, and have experience determining when AI, automation, or data modeling is the right tool.

This sits at a crossroads: some organizations expect every strategist to be AI-literate rather than creating a siloed specialist role. Critics argue AI is a tactical tool to inform strategy, not deliver strategy itself. However, agencies actively hiring report significant client demand for dedicated expertise, with some expanding AI strategy departments specifically due to requests. The pattern mirrors digital strategists a decade ago—initially specialized, eventually distributed. For startups: if you're under 50 people, distribute AI literacy across your product and marketing teams. If you're scaling with enterprise clients asking for AI capabilities, consider a dedicated role.

Effective AI strategists bridge the anxiety gap by reframing AI from job replacement to augmentation. When employees hear 'process automation' and 'AI taking mundane tasks,' service-based workers naturally worry about displacement. The strategist's role includes understanding people behind the process, their frustrations, and friction points—then repositioning AI as enabling higher-order thinking through human-machine collaboration. Real example: McDonald's China used Azure AI and GitHub Copilot with strong change management, increasing employee AI transactions from 2,000 to 30,000 per month. The key is empathy combined with clear outcome measurement that shows how AI creates new opportunities rather than just eliminating tasks.

Start with the NIST AI Risk Management Framework as your foundation and implement it within 30-60 days. Your immediate priorities: complete an AI use case inventory across all departments, establish lightweight AI review gates for new deployments, create AI risk registers and model cards for documentation, and implement data controls including consent management, licensing verification, retention policies, and access restrictions. Assign clear ownership to each use case (product owner, data/ML lead, security lead, legal/compliance representative) and establish when Ethics Committee or AI Steering Committee review is required. Schedule quarterly briefings for leadership and board members to maintain governance visibility.

The most common failures include pilot purgatory (running endless experiments without production deployments), misaligned objectives (AI for AI's sake rather than tied to measurable business outcomes), tool sprawl without workflow integration (adopting point solutions that don't talk to each other), and weak data governance where risk teams scramble after deployment instead of being engaged from day one. Another critical mistake: setting unrealistic expectations by overestimating current AI capabilities and treating AI like plug-and-play software instead of a capability requiring trust, context, and iteration. Avoid these by tying AI to specific outcomes, prioritizing a few high-ROI use cases, building governance early, and instrumenting value measurement from the start.

Use AI consulting if you're pre-Series C with under $40M ARR, need specialized expertise for 3-6 month projects, or require less than 120 hours/week of AI work. Consulting delivers faster time-to-first-result (typically 3 weeks vs 7 months for in-house) and includes multi-person expertise at lower Year 1 cost ($240K-$380K vs $1.4M for full team). Hire full-time if AI is your core product differentiation, you need 3+ dedicated people, you're in an AI talent hub with competitive compensation, or you have a 12+ month sustained roadmap. The strongest approach: start with contractors to validate business cases and build internal confidence, then transition to full-time hires as AI becomes core to operations.

A strong AI strategist job description should outline these core responsibilities: develop and implement AI strategies aligned with business objectives, collaborate with departments to integrate AI solutions into operations, lead AI project development and monitor performance, analyze data and industry trends to identify AI opportunities, and oversee compliance with data protection regulations and ethical AI practices. Required qualifications typically include Bachelor's or Master's in AI/Data Science, 3+ years experience in AI strategy and project management, proficiency in AI technologies including machine learning frameworks, excellent communication and leadership skills, and strong analytical problem-solving abilities. The role should report to the CTO and focus on strategically planning initiatives to align with business goals while facilitating collaboration between departments.

Start by defining clear purpose and scope—what specific problems will the agent solve and what tasks will it handle autonomously versus with human oversight. Key steps: collect and prepare training data, choose the right machine learning model (pre-trained like GPT or custom-built), set up your environment with necessary frameworks (TensorFlow, PyTorch, Dialogflow, or Rasa), split data into training and testing sets, and design the agent's reasoning and planning capabilities. Foundational components include a large language model for reasoning, a memory system, an action interface for tool use, and a mechanism for perceiving its environment. Critical challenges to address: ensuring reliable performance, managing complex multi-step tasks, debugging autonomous behaviors, and addressing safety and ethical concerns. This is an iterative process requiring continuous refinement based on testing results.

For pre-seed to seed stage (under 10 people), your first AI hire should be a versatile AI/ML engineer who can both build and strategize, reporting directly to a technical founder. Don't hire a specialized AI strategist until you have at least 15-20 people and multiple departments requiring coordination. Alternative approach: use AI tools as your 'first hire' before bringing on human talent—AI can take on analyst, designer, and strategist roles to bridge early capability gaps and help you move from idea to execution without waiting for headcount. Start with ChatGPT for job descriptions and process documentation (free, immediate results), then add Dover for candidate screening, then Tactiq for interview documentation as you scale. This lean approach lets you validate AI direction and build internal capability before committing to full-time specialized roles.

Track outcome-based metrics, not activity metrics. Primary KPIs include: number of new projects completed that wouldn't have been possible without AI (innovation enablement), measurable impact on specific business metrics (revenue growth, conversion rates, customer acquisition cost), time-to-production for AI initiatives (speed from concept to deployment), and cross-departmental AI adoption rate (measuring how well silos are being broken down). Secondary metrics: AI governance compliance rate (percentage of deployments following established frameworks), employee AI literacy improvement (measured through training completion and usage), risk mitigation value (quantified compliance improvements and incident reduction), and strategic value delivery (competitive capabilities gained). Avoid weak metrics like 'efficiency gains' or 'time saved'—focus instead on what value those gains delivered to the business.

Establish a lightweight governance mechanism rather than heavy bureaucracy—this can be a technical board, council, or even a single deeply embedded person. Implement these 30-60 day priorities without creating bottlenecks: adopt NIST AI RMF as your framework, create simple AI risk registers (one page per use case), establish clear approval structures (assign product owner, data/ML lead, security lead, legal rep to each project), and define when committee review is required versus when teams can move autonomously. Build governance around five key principles: fairness, transparency, accountability, privacy, and security. The trick is turning principles into action through clear policies and continuous monitoring, not creating approval layers that slow deployment. Engage legal, security, and risk teams on day one to define approved data sources, redaction policies, and human-in-the-loop checkpoints before scaling beyond pilots.

Not yet, but AI tools can augment early-stage strategic work before you're ready to hire. AI can help with analyst tasks (processing data to identify opportunities), designer tasks (prototyping solutions), and basic strategist work (generating frameworks and documentation). However, AI cannot replace the human judgment required for: breaking down organizational silos between departments, building empathy with teams anxious about automation, making ethical decisions about deployment contexts, navigating complex stakeholder dynamics, and tying AI capabilities to nuanced business outcomes. As one example shows, AI-assisted teams can bypass traditional hiring bottlenecks and move from idea to execution faster, but strategic oversight still requires human expertise—especially for determining when to deploy generative versus agentic AI, establishing governance that protects confidential data, and measuring outcome value rather than just efficiency.