#45: McKinsey reinvents consulting model with 12,000 AI Agents

In partnership with

📢 TOP Headlines

McKinsey Deploys 12,000 AI Agents

McKinsey has quietly implemented 12,000 AI agents across its consulting operations—marking a bold shift from human-heavy processes to a hybrid model of human plus agent-driven delivery. These agents now draft reports, build PowerPoint decks, verify logical arguments, summarize interviews, and take notes, operating at scale behind the scenes.

What This Signals:

• Productivity powered by agents: The firm has trimmed staff from ~45,000 to ~40,000, yet continues to hire. With fewer junior employees needed for routine work, McKinsey is redefining its client teams: a manager and 2–3 consultants + AI agents now replace formerly large groups.

• Agents enforcing the McKinsey voice: One of the most-used bots adopts the firm’s signature “sharp, concise, clear” style. Another reviews the logic flow of arguments, suggesting agents are being used to uphold brand and quality consistency.

Takeaways for Leaders:

• High-value human + agent hybrid is the new staffing model: Junior roles focused on repetitive tasks are shrinking, but human leadership and judgment become even more essential.

• Agent output must reflect brand identity: Voice, tone, and structural quality are now agent-level product requirements - not just model accuracy.

• Outcome-based consulting is agent-enabled consulting: With 25% of work tied to outcomes, agents make execution measurable, scalable, and monetizable.

McKinsey’s approach flips the narrative: AI isn’t about job cuts - it’s about restructuring work around scalable intelligence. Agents shoulder the foundation so humans can focus on creativity, leadership, and nuanced decision-making.

If your startup or team is building agent tools, it’s time to think about:

• Building voice-embedded, brand-aligned agents

• Structuring human+agent teams with complementary skills

• Shifting toward performance and outcome-based pricing models

Perplexity’s $34.5B Chrome Bid

AI startup Perplexity AI, valued at approximately $18–20B, is grabbing headlines with an unsolicited $34.5B all-cash offer for Google’s Chrome browser. Here's what’s really going on and why it matters in the agent era.

What Perplexity Is Doing (and Saying)

• The offer arrives amid antitrust pressure on Google, suggesting strategic positioning if regulators force a divestiture.

• The proposal includes a pledge to:

  • Keep Chromium open-source

  • Inject $3B into Chrome development

  • Respect user settings (defaults, UX, support) for the next 100 months

• Analysts view the move as a PR masterstroke, designed to disrupt public perception and define Perplexity’s ambition—and it comes as the startup ramps up funding.

Reddit Buzz

The Reddit thread on this news is filled with skepticism:

This isn’t about Perplexity paying Google for Chrome, Google almost certainly won’t sell. It’s a statement: Perplexity isn’t just building agents - it wants to own the browser layer they run on. In this phase of AI, platform position may matter as much as model performance.

⚔️ AArena: All-in-One AI Workspace

Get Sh*t Done with AI.
Bring your task. Get best results.

TOP 5 AI (Aug 14):

1. Grok 4

2. Gemma 3 4b

3. QwQ 32B

4. ChatGPT-4o

5. Qwen3 32B

👉 Enter the Arena
👉 Battleboard

📡 TOP Agent Signals

• Factory 1.5 introduces a fully redesigned Session interface

• Salt Security study finds 62% of people believe AI agents are easier to deceive

• Capitalize launched the first AI agents for commercial real estate financing.

• Apiiro announced AutoFix Agent.

• Superagent AI announced first fully autonomous AI insurance agent.

• Egnyte unveiled Agent Builder

• Alibaba International released “Accio Agent”

• Apexon unveiled AgentRise.

  - - -

• A new preprint surveys the emerging field of self-evolving AI agents, which continuously adapt their behavior via feedback loops. The paper bridges static foundation models and adaptive agentic systems, proposing a unified framework for agents that learn and evolve over time to handle dynamic real-world tasks

• ArXiv survey compares three approaches to AI agent registries, which manage discovery and identity of autonomous agents. It examines MCP’s GitHub-based metaregistry, the decentralized Agent-2-Agent (A2A) framework of JSON Agent Cards, and NANDA’s blockchain-backed AgentFacts, analyzing how each supports secure, interoperable ecosystems of AI agents

• An arXiv paper introduced “Agent Lightning,” a reinforcement-learning framework for training LLM-based agents. It decouples agent execution from training and applies hierarchical RL to agent workflows, enabling RL fine-tuning of agents built with diverse frameworks.

• A preprint proposes “AgentOps”, a structured operations framework for agent systems. It first categorizes the types of anomalies that occur within and between AI agents, then defines four stages of AgentOps – monitoring, anomaly detection, root-cause analysis, and resolution – as essential practices to stabilize and secure agentic applications.

• A research paper introduces AIAP, a no-code platform that lets non-experts build AI-driven workflows in natural language. AIAP uses a coordinated multi-agent system to decompose vague user instructions into modular steps behind a visual interface; a user study showed this approach significantly improves users’ ability to design AI services by reducing complexity and helping them build effective agentic programs.

🗺️ AI Agents Landscape Map

We’re mapping thousands of AI Agents. Find your next AI agent today — explore thousands of options on our interactive Landscape Map.

👉 Explore the Map