AI Consultant — Start With the Problem, Not the Tool
An AI consultant helps an organisation determine what to build, what to buy, what to automate, what not to automate, and how to move from an AI idea to something the business can actually use. A company does not need an AI consultant because AI is fashionable — it needs one when it has a business problem that AI might improve. This work sits at the intersection of AI product strategy, technology development, software delivery and commercial decision-making.
- AI Strategy and Discovery (1–3 weeks): You know AI could help but do not yet know exactly what to build. Designed to stop an organisation spending significant money building the wrong thing. Covers business objectives and existing workflows, operational bottlenecks and available data, current technology and ai opportunities, ai risks and expected business value, build-vs-buy options.
- AI Architecture and Solution Advisory (2–6 weeks): The use case is reasonably clear, but you need help determining how the system should actually work — so it survives beyond the demo. Covers llm selection and ai application architecture, rag architecture, vector search and data pipelines, ai agents and tool calling, api integrations and cloud architecture, security and human-in-the-loop controls, evaluation, monitoring and production considerations.
- AI Vendor and Technology Selection (1–4 weeks): You are comparing AI platforms, LLM providers, AI SaaS products, automation platforms, development partners, RAG or agent platforms, or cloud AI services. Covers functional fit and integration requirements, security, data handling and scalability, model flexibility and vendor lock-in, cost and implementation effort, support and long-term maintainability.
- AI Implementation Advisory (4–12+ weeks): You have selected a direction and need experienced guidance through implementation — alongside an internal engineering team, an external development partner or the Ortem Technologies delivery team. Covers architecture review and requirements clarification, product decisions and ai workflow design, vendor coordination and delivery oversight, evaluation strategy and risk review, production readiness and stakeholder communication.
- Fractional AI Product or Technology Advisory (3–12 months): You need senior AI and technology decision-making but are not ready for a full-time AI executive or CTO. Particularly useful for startups and growing businesses building AI capability. Covers ai roadmap and product decisions, technology selection and vendor management, architecture review and delivery oversight, gtm implications and enterprise customer requirements.
- Deliverables: AI Opportunity Assessment, AI Use-Case Prioritisation Matrix, AI Strategy and Roadmap, Solution Architecture, Build-vs-Buy Recommendation, Vendor Evaluation Matrix, Implementation Roadmap, AI Governance Recommendations.
- Worked example — Fleet management: A fleet-management business was operating across spreadsheets and disconnected tools. Fleet managers were manually tracking vehicle maintenance, fuel, driver assignments, inspections and parts across fleets of roughly 100+ vehicles. Missed maintenance reminders were creating operational problems, fuel costs were difficult to track, and driver-reported issues were handled over email rather than structured workflows. The real problem was not a need for AI — it was fragmented operational information. Approach: Delivery started with structured discovery involving fleet managers, maintenance managers, mechanics and drivers. The workflows were mapped before the product architecture was finalised. Reported results: 8–15% fuel-cost reduction, 40%+ downtime reduction, 3× faster work-order throughput, 142+ vehicles managed per organisation, Five role-based access levels. Figures as reported in the published case study, which records the 8–15% fuel-cost reduction as achieved within 90 days. A related Ortem client testimonial states the platform included an AI anomaly-detection layer, with Praveen involved hands-on throughout.
- Worked example — Fintech: Stock Assist AI was designed around a fragmented research workflow. Users were switching between stock charts, news, technical indicators and an AI assistant that did not inherently understand the market data being viewed. The opportunity was to bring those workflows together rather than create another isolated chatbot. Approach: The product workflow was broken into seven friction points — stock lookup, chart interpretation, news analysis, technical indicators, portfolio comparison, real-time alerts and trade-decision synthesis — then addressed with an agentic architecture where tools are called as part of a single interaction. Reported results: 46 active paying users at peak, 99.8% uptime, Under two-second AI response time, 15,000+ AI queries processed, 98.5% payment success rate. Figures as reported in the published case study. The project had a published value of more than $95,000.
- Build-vs-buy is decided against the business requirement, not a default preference: define the problem, determine whether AI is actually necessary, evaluate existing products, calculate the full economic case, compare ownership and flexibility, then state buy, build, customise or hybrid and explain why.
- What I explicitly do not do: recommend ai simply because ai is available; sell an ai demo as a finished production system; promise autonomous ai where human control is necessary; pretend every company needs a custom model; choose technology based purely on popularity; promise a guaranteed roi before understanding the baseline; take on projects where the strategic objective is unclear.