AI for every role

AI handles the volume side of accounting (categorization, reconciliation, expense coding) at production quality, but it still cannot make the judgment calls on complex tax positions or audit opinions that define the profession; firms that deploy it for data entry gain 10+ hours per week per staff member, and firms that trust it for tax strategy will get burned.

AI cuts first-draft time on reference docs and how-to guides by roughly half, but it still invents parameters, flags, and API fields that do not exist in your codebase.

Consulting firms that use AI widely in service execution show 17.9% EBITDA versus 6% for firms that do not (Rocketlane 2026 benchmark, 509 organizations), but the edge comes from client-facing automation (proposals, deliverables, analysis), not internal productivity tools; solo consultants get the biggest lift from AI because they have no junior staff to delegate to.

AI drafts and triages tickets fast enough to cut first response time in half, but it still mishandles or wrongly closes at least 1 in 10 refund and account access requests without flagging the miss.

AI accelerates the production phase of design by 3-5x (wireframes, asset generation, icon sets, prototype iteration), but it cannot do the research and problem-framing that separates good design from decoration; 91% of designers use AI weekly, and the ones who gain the most treat it as a production tool, not a design partner.

87% of marketers use AI in recurring workflows and most still use it as a content mill; the bigger payoff is in predictive audience segmentation and campaign analytics, where AI-assisted teams see a 27% lift in email open rates, but brand voice consistency remains unsolved and every draft still needs a human editor who knows the brand.

AI now does the writing half of product management well (PRDs, summaries, updates) and still cannot do the deciding half (prioritization, tradeoffs, saying no); a PM who treats it as a drafting engine gains a day a week, and one who treats it as a decision engine ships worse roadmaps faster.

AI now writes the code half of data engineering well (SQL, dbt models, pipeline boilerplate) and stays dangerous on the data half: a wrong transformation runs clean and returns plausible numbers, so every AI-written query needs a validation step the tools do not provide.

AI literature review tools cut discovery time roughly in half and structured extraction tools like Elicit turn weeks of manual screening into hours, but researchers who skip manual citation verification introduce errors that peer reviewers catch, and fabricated references remain the highest-profile failure mode in academic AI use.

AI cuts demo-prep and RFP-response time by roughly half, but it still invents technical specs, so a sales engineer must verify every claim before a prospect sees it.

AI tools cut UX research synthesis time roughly in half, but they still misattribute a quote to the wrong participant in more than 1 of every 5 sessions checked.

AI cuts first-pass policy review time for compliance officers by roughly half, but in testing it missed 2 of 5 planted conflicting-clause errors across ChatGPT Enterprise, Claude, and Harvey, so nothing ships without a human sign-off.

AI cuts recruiter sourcing and screening time by roughly half, but every HireVue score and every AI-drafted outreach message still needs a human read before it reaches a candidate.

AI tools cut the drafting time on comps memos and market summaries by half, but none of them will catch a broken formula reference in your own model, so you still check every number by hand.

Legal AI tools can draft contracts and summarize case law faster than junior associates, but hallucinated case citations remain a career-ending risk that no tool has fully solved; CoCounsel and Lexis+ with Protege reduced hallucination rates to 4-7% on citation tasks, which is an improvement, and still far too high for a profession where one fake citation means sanctions.

AI does first-pass document review and summarization well enough to save a paralegal several hours a week, and remains unfit for anything citable: fabricated authorities are still routine in 2026, so every case, statute, and clause reference must be verified in Westlaw or Lexis before it leaves your desk.