Skip to content
GFB CapitalFR

AI tools for consultants in 2026: research findings, comparison and practical guidance

Productivity & AI · · 8 min

Situation, complication, recommendation

Situation. Consulting is one of the professions most exposed to generative AI: research, synthesis, writing and formatting account for a large share of time. Major firms have developed their own tools: according to figures released by the firm, more than 75 % of McKinsey employees use its internal Lilli platform each month [5], and around 40 % of BCG junior consultants use its Deckster presentation tool each week [4].

Complication. The range of products has multiplied, prices and features change every quarter, and gains are not automatic: used poorly, AI undermines the accuracy of analysis.

Recommendation. Build a short combination of tools, selected by use case, starting with the office suite already in place, and govern it with five simple rules.

1. What the studies measure

Harvard Business School and BCG (2023). In the study “Navigating the Jagged Technological Frontier” (Fabrizio Dell'Acqua, Ethan Mollick, Karim Lakhani et al.), a controlled, pre-registered experiment involving 758 BCG consultants, around 7 % of its consultants, those with access to GPT-4 completed 12.2 % more tasks, 25 % faster, with quality judged more than 40 % higher on tasks suited to AI (product ideation, launch planning) [1]. The consultants who initially performed least well improved most (+43 %, compared with +17 % for the best performers). But on a task designed to fall outside this “frontier” — the “jagged technological frontier” described by the authors — (a business case combining numerical data and interviews), assisted consultants were 19 percentage points less likely to find the correct answer [1].

Effect of GPT-4 on BCG consultants’ work, tasks suited to AI (difference from the group without AI, in %)

Source: Dell'Acqua et al., Harvard Business School, 2023, n = 758. — Chart redrawn by GFB Capital from published data.

Procter & Gamble (2025). The study “The Cybernetic Teammate” (Fabrizio Dell'Acqua, Ethan Mollick, Karim Lakhani et al., NBER), involving 776 P&G professionals working on real product innovation topics, shows that one person assisted by AI reaches the level of a two-person team without AI (+0.37 standard deviations versus +0.24), with 12 % to 16 % of time saved [2]. Teams equipped with AI produce the most top-tier solutions, and AI helps both technical and commercial specialists propose more balanced solutions.

Companies. In McKinsey’s global survey “The state of AI in 2026”, published in August 2026, 80 % of respondents report that AI has improved their individual productivity [3].

Two caveats: these experiments were conducted with GPT-4 and GPT-4o, which are less capable than current models; and their authors observe that assisted outputs are more homogeneous, which may reduce the diversity of ideas [1].

2. Comparison of the main tools (October 2026)

AI tools useful to consultants: uses, strengths, limitations and public prices (USD excluding tax per user per month, October 2026)
CategoryToolMain useStrengthWatchpointIndicative public price
Integrated assistantMicrosoft 365 CopilotWriting, analysis and synthesis in Word, Excel, PowerPoint, Outlook and Teams, using company data; Researcher and Analyst agentsNo change of tool; access to emails, files and meetingsRequires a Microsoft 365 licence; value depends on the quality of internal data30 (enterprise plan, annual commitment); 21 (Copilot Business, up to 300 users) [6]
Integrated assistantGemini in Google Workspace and NotebookLMAssistant in Gmail, Docs, Sheets, Slides and Meet; NotebookLM for querying a closed set of documentsIncluded in the subscriptionLimited features in the Business Starter planIncluded from 14 (Business Standard, annual commitment) [7]
General-purpose assistantChatGPT BusinessWriting, data analysis, deep research, agent modeVersatility, broad ecosystemMinimum of two users; Enterprise plan priced on request20 (annual); 25 (monthly) [8]
General-purpose assistantClaude (Team plan)Analysis of long documents, writing, structured reasoning; Microsoft 365 and Google Workspace integrationsProcessing large documentsMinimum of two users; Enterprise plan partly usage-based20 (annual); 25 (monthly) [9]
ResearchPerplexity Enterprise ProWeb research with citations, research reportsVisible, verifiable sourcesResearch tool, not a task execution tool34 (annual); 40 (monthly) [10]
Due diligenceSpecialist platforms (AlphaSense, Hebbia…)Market intelligence, analysis of large document volumesDesigned for investment and due diligenceHigh costOn request
PresentationsGammaFirst draft of a presentation from an outline or documentSpeedLess precise layout control than PowerPoint8 to 10 (Plus); 15 to 20 (Pro) [11]
Presentationsthink-cellConsulting charts in PowerPoint (waterfalls, Marimekko, Gantt) linked to ExcelIndustry standardNot a generative AI toolOn request
MeetingsGranolaMeeting notes captured from computer audio, with no visible bot in the callDiscretion, enriched notesInform participants that notes are being taken14 (Business); 35 (Enterprise) [12]

indicative public prices recorded in October 2026 (US prices, excluding tax); they change frequently and may differ in Europe. Reliability: medium — cross-checking of consistent secondary sources, to be confirmed on each vendor’s website.

3. Which combination should you choose?

The principle: start with the tool integrated into the suite already in use, then add only one tool for each unmet need.

  • Independent consultant or very small firm: a general-purpose assistant (ChatGPT or Claude, around 20 USD per month on an individual plan), Perplexity Pro (20 USD), Granola (14 USD) and Gamma (8 to 10 USD), totalling around 62 to 64 USD per month (GFB Capital calculation based on public prices).
  • Firm using Microsoft 365: Copilot for work in documents, emails and meetings, a general-purpose assistant for analysing long documents, and think-cell for charts.
  • Firm using Google Workspace: Gemini and NotebookLM, already included, supplemented by a general-purpose assistant and a sourced research tool.
  • Investment and due diligence teams: add a specialist platform for analysing data rooms and market intelligence.

4. Five rules for use

  1. Know the frontier. Delegate first drafts, synthesis and formatting; retain judgement, key figures and the recommendation. Verify every figure and every source: outside the frontier, AI costs 19 points in correct answers [1].
  2. Protect client data. Use professional plans, check clauses preventing data reuse for model training and the hosting location, and obtain the client’s agreement when processing their documents.
  3. Inform before recording. Notify participants before any meeting transcription, as required by the GDPR.
  4. Preserve diversity of ideas. Compare several sources or models, and retain time for thinking without AI.
  5. Measure. Track time saved and the quality of deliverables, and reinvest the time freed up in analysis and the client relationship.

5. Our takeaways

GFB Capital uses AI agents daily for research, analysis and the production of deliverables. The studies confirm our experience: AI significantly accelerates collection, structuring and formatting, but the value of an assignment still lies in diagnosis, judgement and execution alongside management.

Our AI R&D

Limitations. The cited experiments concern specific tasks and models from 2023–2024; firms’ figures are self-reported; prices are public prices subject to change.

Sources and credits

References consulted on 31 October 2026.

  1. [1] Fabrizio Dell'Acqua, Edward McFowland III, Ethan Mollick, Hila Lifshitz-Assaf, Katherine C. Kellogg, Saran Rajendran, Lisa Krayer, François Candelon and Karim R. Lakhani, « Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality », Harvard Business School Working Paper no. 24-013, September 2023 — experiment conducted with the BCG Henderson Institute involving 758 BCG consultants. Randomised experiment — high reliability.
  2. [2] Fabrizio Dell'Acqua, Charles Ayoubi, Hila Lifshitz, Raffaella Sadun, Ethan Mollick, Lilach Mollick, Yi Han, Jeff Goldman, Hari Nair, Stewart Taub and Karim Lakhani, « The Cybernetic Teammate: A Field Experiment on Generative AI Reshaping Teamwork and Expertise », NBER Working Paper no. 33641, March 2025 — 776 Procter & Gamble professionals. Randomised experiment — high reliability.
  3. [3] Dan Tinkoff, Lieven Van der Veken and Michael Chui, with Tara Balakrishnan, « The state of AI in 2026: On the road to ROI », McKinsey & Company / QuantumBlack, AI by McKinsey, 25 August 2026 — 1 719 participants, 97 countries. Primary source — high reliability (self-reported data).
  4. [4] Lakshmi Varanasi, « Inside the AI boom that's transforming how consultants work at McKinsey, BCG, and Deloitte », Business Insider, 27 April 2025 — figure reported by BCG. Medium reliability.
  5. [5] Entrepreneur, « McKinsey Is Using AI to Create PowerPoints and Take Over Junior Employee Tasks », June 2025 — figure reported by McKinsey. Medium reliability.
  6. [6] Microsoft 365 Copilot prices: recorded on 14 October 2026 by VDF AI, consistent with other pricing guides. Secondary source — medium reliability.
  7. [7] Google Workspace prices: recorded on 27 October 2026. Secondary source — medium reliability.
  8. [8] ChatGPT Business prices: October 2026 pricing guide, consistent with other sources. Secondary source — medium reliability.
  9. [9] Claude Team prices: recorded in September–October 2026. Secondary source — medium reliability.
  10. [10] Perplexity prices: pricing guide updated on 8 October 2026. Secondary source — medium reliability.
  11. [11] Gamma prices: recorded in June 2026. Secondary source — medium reliability.
  12. [12] Granola, “Granola pricing: Plans, features, and ROI calculator for teams”, 2026. Vendor source — high reliability for the displayed prices.

The cited figures, analyses and frameworks remain the property of their authors and publishers, to whom we refer readers for their full methodology. The cited trademarks belong to their respective owners. GFB Capital is not affiliated with any of the organisations cited.

Contact us
← All articles