- First Principles Marketing
- Posts
- Centaur State of Play: Why Human+ AI Beats Either Alone
Centaur State of Play: Why Human+ AI Beats Either Alone
Still staging a cage‐match between people and robots? That ended in the ’90s. The 2025 edge belongs to the centaur – a strategist riding a herd of specialised AI agents.
Suyash Chaudhari
June 04, 2025
TL;DR (Because You’re Busy)
Stop debating man vs. machine. Pair them.
Start with one annoying task. Give it to an agent; judge the delta.
Stay the strategist. Let software hammer the tactics.
Marketing isn’t magic; it’s physics. And in 2025, the longest lever is a centaur.
📚 First-Principles Playbook: 5-Part Series
1. Leverage > Labor
2. Centaur Advantage ← you’re here
3. Taste Is Currency
4. Velocity Loops Win
5. Small Teams, Enterprise Power
The Pattern: Machines Tackle Tactics, Humans Own Strategy
1997 - Deep Blue proved computers could brute‑force their way past grandmasters. Kasparov’s revenge plan was counter‑intuitive: work with the silicon, not against it.
1998 - Advanced Chess
Format: one human + one computer on each side.
Outcome: Kasparov’s lifetime library of tactical patterns suddenly mattered less, the machine spat perfect tactics; he focused on the long‑game. Result? Upsets galore.
"Human creativity was even more paramount under these conditions, not less.
My advantage in calculating tactics had been nullified by the machine.”
Mid‑2000s - Freestyle Chess
Format: any mix of humans & computers, even multiple engines.
Outcome: Two amateur club players + three laptops not only demolished Hydra, then the world’s most powerful chess computer, but also crushed grandmasters + computers teams.
Key skill wasn’t chess brilliance; it was orchestrating several engines, probing lines, and synthesising a strategy.
The humans on the winning team were the best at "coaching" multiple computers on what to examine, and then synthesizing that information for an overall strategy. Human/Computer combo teams-known as "centaurs” - were playing the highest level of chess ever seen.
"You hear people that talk about their job now is to assign work to a bunch of agents, look at the quality, figure out how it fits together, give feedback, and it sounds a lot like how they work with a team of still relatively junior employees”
Notice the rhyme: both Kasparov’s centaur captain and Altman’s agent-manager spend their day prioritising, orchestrating, and refining. They don’t brute-force every move; they decide what matters, direct silicon muscle to dig, then tighten the output. Same blueprint, different board.
The competititve edge shifts to the human who can coach a fleet of AI tacticians.
See where we’re going? Swap rooks for revenue ops, or SaaS roadmaps, or pricing experiments, and you’ve got 2025. LLM agents spin 50 ad‑variants before you sip coffee, but they still need a human to decide which hill to take and why.
2025: Your New Teammates Are Digital Specialists
Talk to five SaaS operators and you’ll hear two very different dreams.
Dream #1: “One click, replace the hire.”
Dream #2: “Give me a tireless co-pilot that plugs into my workflow.”
Spoiler: Dream #2 is winning budgets right now.
Alina Vandenberghe nails the vibe:
I can see a future in which core marketers, humans who setup the narrative and own the P&L, are surrounded by AI agents that are digital specialists across various paths and become their executive arms at scale.
Translation: keep the compass, let the bots row. The agent landscape now splits into two tribes: ‘Full Hire’ replacements chasing total autonomy and Copilots that slot into human workflows. Guess which one is getting traction?
Autonomy Level | Example Tool | Adoption Snapshot | Take‑home |
|---|---|---|---|
Full “hire” replacement | Devin: autonomous software engineer (update - “collaborative AI teammate” now😅) | New pricing plan: from only $500 plan to $20/month in Apr 2025 | Hype high, penetration tiny |
Copilot mode | Cursor IDE: pairs in‑editor | 7M devs, >$200 M ARR by Mar’25 the fastest SaaS ramp ever. (sacra.com, creatoreconomy.so) | Human stays in the loop; efficiency gains for devs. |
Pattern: tools that augment humans are winning mind‑share and wallets faster than those that promise to replace us outright.
Why the Centaur Wins
In the end, Kasparov did figure out a way to beat the computer: by outsourcing tactics, the part of human expertise that is most easily replaced.
That chess epiphany carries straight into SaaS ops.
Offload any step where “right answer” = deterministic grind; keep the calls that hinge on judgment, context, and taste.
Domain | Human | AI Agent | Combined Impact |
Marketing | Defines narrative & positioning | Generates, repurposes, and A/B tests assets | Narrative coherence and variant explosion |
Customer Success | Relationship judgement & escalation | Remembers every ticket, click & call | Faster resolutions & higher CSAT |
Software Dev | Architectural insight | Autocomplete, refactor, test generation | Dev velocity >2× |
Key takeaway: Outsource precision; keep perspective.
Playbook: Build Your Centaur Stack
Audit your “tactics tax.” List repetitive chores stealing strategic hours - manual lead-routing, CRM enrichment, stitching product‑usage data into board slides, weekly churn snapshot wrangling.
Insert an agent where error‑cost is low & feedback is fast. Think creative iteration before live budget re‑allocation.
Gate releases. Human sign‑off until the agent’s error‑rate < 2 %.
Instrument everything. Treat agents like interns - track PRs, uptime, ROI.
Promote or fire quickly. If an agent stagnates after three sprints, demote it to a script and move on.
If your “AI agent” is really just a renamed Zap called automation_v2, you’ve automated clicks, not decisions.
Mind the Blind Spots
Hallucination risk: LLMs still invent facts; chain two hallucinations and you have brand damage at machine speed.
Data entropy: Most orgs haven’t prepped data pipelines for gen‑AI; the agent can’t fix what the CRM broke.
Security & misalignment: Greater autonomy raises proportional risk.
Put simply: centaurs still need guardrails. As Yoshua Bengio’s new LawZero initiative argues, we may soon need “AI watchdogs” to keep other agents honest.
Manageable? Absolutely. Big‑tech agent builders - Microsoft Copilot Studio, Google Vertex AI Agent Builder, AWS Bedrock Agents - now bake in identity isolation, deterministic guardrails, and plug‑and‑play connectors so your centaur doesn’t gallop off the rails.
What’s Next? Three Predictions
Proactive > Reactive: Agents will shift from “ask‑and‑respond” to “monitor‑and‑act,” surfacing decisions before you open Slack.
Skill arbitrage flips hiring: Founders will hire for taste and judgement; agents will supply executional bandwidth.
The Centaur League: Communities will spring up where marketers agent‑playbooks the way chess players once traded openings.
Net-net: the marketers who learn to captain agents, not compete with them, will write the playbooks everyone else Googles next year.
Was this useful? Hit reply or forward to a friend who’s still solving tactical puzzles by hand.
About Me
I’m Suyash - badminton junkie, ex‑GroupM ad‑ops grunt, first marketer at a B2B SaaS startup, and creator of Otto, the paid‑search autopilot.
My mission: think, so you can click less.
Let’s build leverage together.
Reply