The first step in working productively with AI is understanding what a practitioner's own work actually consists of. Most professionals describe their work through their role (consultant, lawyer, analyst, broker) rather than through the specific tasks the role involves. Role descriptions are useful for business cards and employment contracts. They are less useful for deciding where AI can help, because a role is too abstract to map onto specific AI capabilities.
A more useful approach is to decompose the role into the tasks that actually consume a practitioner's time. Research. Drafting. Analysis. Communication. Administration. Meeting preparation. Document review. Every professional job breaks down into a specific combination of these and similar categories, and the specific combination varies substantially across roles within the same broad profession. A corporate lawyer's days are different from a litigator's days. A management consultant's days are different from a financial-advisory consultant's days. A listing-side residential agent's days are different from a buyer-side agent's days. The task mix determines where AI can help, and the starting point for productive AI use is an honest inventory of the tasks that actually make up the work.
The inventory exercise is straightforward. Write down the ten most frequent tasks in the week. For each, note how often it happens (daily, weekly, monthly, occasional) and what kind of work it actually involves (is it mechanical production, is it analytical, is it judgment-heavy, is it relationship-driven). This produces a map of the work that can then be evaluated for AI opportunities.
A useful screening framework for each task asks three questions. The first is whether the task is repetitive in structure, meaning it takes roughly the same shape each time even when the specific content differs. Drafting a weekly client update, preparing a standard case summary, writing quarterly performance reviews. The second is whether the task relies on information that already exists in digital form, such as documents, emails, prior work product, databases, or notes, rather than on things only the practitioner knows. The third is whether the practitioner would still review the final product before it reaches its audience, which ensures that AI is augmenting rather than replacing professional judgment. If the answer to at least two of the three questions is yes, the task is typically a strong candidate for AI support.
Tasks that commonly qualify include drafting initial versions of memos, emails, and reports. Turning bullet points or meeting notes into structured documents. Preparing first drafts of presentations. Organising notes from meetings or interviews into clean records. Producing checklists, templates, and standard operating procedures from existing process documentation. Summarising long documents. Extracting structured information from unstructured sources. These are the tasks where AI assistance produces the most immediate value, and the tasks where the learning curve is easiest for a practitioner starting to work with AI.
Tasks that typically do not qualify for AI assistance include relationship-driven work, judgment calls on edge cases, decisions that require accountability, and tasks that depend on information the practitioner has that is not in digital form. A client conversation about a difficult situation, a judgment call on whether to accept a deal term, a decision about whether a particular recommendation fits the specific context. These fall outside AI's scope. The tasks in this category are where the practitioner's professional value is most concentrated, and AI use should free up time for them rather than attempting to replace them.
The important principle is that practitioners should not force AI into every part of their work. Force-fitting AI onto tasks where it does not belong produces poor output and wasted effort. Applying AI thoughtfully to the tasks where it substantively helps produces substantial time savings and often improved output quality. The task inventory is the foundation for making that distinction deliberately. A practitioner who knows which of their ten most frequent tasks are good candidates for AI has a map they can start working from. A practitioner who has not done the inventory tends to either under-use AI (because they have not identified the opportunities) or over-use AI (because they apply it where it does not fit and produce disappointing results that sour them on the tools generally).
The inventory is also worth revisiting periodically. The AI tools themselves continue to improve, and tasks that were not good candidates a year ago may be good candidates now. The practitioner's role may evolve, and the task mix may shift. The practice of maintaining an active view of what the work is composed of is itself a professional discipline, and one that pays off in ways that extend beyond AI use.