{"stable_id":"0dc0a03188456b2e","slug":"bcg-gpt4-jagged-frontier","company":"Boston Consulting Group","workflow_name":"GPT-4 integrated into individual consulting task flows","function_code":"engineering","pattern_codes":["creator_to_judge"],"changed_assumption":"That access to a general model improves all apparently similar knowledge tasks.","evidence_strength":"verified","publication_tier":"showcase","freshness":"current","reviewed_at":"2026-08-23","updated_at":"2026-08-22","source_quality_summary":"2 independent; publication outcomes are verified.","caveat_summary":"Experimental tasks approximated consulting work but were not client delivery; the negative out-of-frontier result is as important as the positive averages.","freshness_caveat":null,"workflow_summary":{"before":[{"actor":"Consultant","action":"Analyze source materials, generate ideas, and draft recommendations unaided by GPT-4.","actor_type":"control"},{"actor":"Consultant","action":"Check coherence and submit the work.","actor_type":"control"}],"hinge":"That access to a general model improves all apparently similar knowledge tasks.","after":[{"actor":"Consultant with GPT-4","action":"Delegate or interleave ideation, analysis, and drafting with the model.","actor_type":"system"},{"actor":"Consultant","action":"Verify, revise, and submit model-assisted work.","actor_type":"system"}],"decision_mode":"shared","decision_marker":"Human moves from creator to judge"},"before":[{"order":1,"actor":"Consultant","action":"Analyze source materials, generate ideas, and draft recommendations unaided by GPT-4.","handoff_to":"Deliverable","control":"Professional judgment"},{"order":2,"actor":"Consultant","action":"Check coherence and submit the work.","handoff_to":"Evaluator or client","control":"Human accountability"}],"after":[{"order":1,"actor":"Consultant with GPT-4","action":"Delegate or interleave ideation, analysis, and drafting with the model.","handoff_to":"Consultant","control":"User prompting and task choice"},{"order":2,"actor":"Consultant","action":"Verify, revise, and submit model-assisted work.","handoff_to":"Evaluator or client","control":"Human final decision"}],"decision_rights":"The model proposes content and analysis; the consultant owns verification and the submitted answer.","exception_path":"For tasks outside the model's capability frontier, consultants must solve independently or validate against source evidence rather than accept the model's confident answer.","removed_work":["Some first-pass ideation, analysis, and drafting on tasks inside the model's capability frontier"],"outcomes":[{"metric":"Task completion speed and correctness by task type","baseline":"Randomized no-AI control consultants","result":"Inside the frontier: 25.1% faster, 12.2% more tasks, over 40% higher quality; outside the frontier: 19 percentage points less likely to be correct","period":"2023 preregistered experiment","scale":"758 BCG consultants, about 7% of individual-contributor consultants","attribution_caveat":"Experimental tasks approximated consulting work but were not client delivery; the negative out-of-frontier result is as important as the positive averages.","evidence_label":"verified"}],"executive_lesson":"The operating control is task classification, not generic prompt training; users need a way to recognize when the model is outside its frontier.","anti_pattern":"Applying the positive average to all knowledge work or assuming prompt training eliminates frontier risk.","questions_for_leaders":["How will users identify out-of-frontier tasks?","What evidence must be checked before submission?"],"collections":["negative-results"],"bundle_version":"1.0.0","bundle_fingerprint":"sha256:c23c6cc2b88153d008ea8fda928f632ce0011fc2d4c5036672a16e5d895bab93","canonical_url":"https://brianletort.ai/transformations/bcg-gpt4-jagged-frontier"}