Bots with Benefits 🤖
When AI Agents Take the Wheel
Hello
Love, Death & Robots is one of my favourite Netflix shows. Short episodes, wild ideas, and always a glimpse of futures where machines take over. In DeFi, we’re living through one such version, except the robots here are driving money.
In the late 2000s, financial advisors started losing sleep over a new invention: the robo-advisor. For a fraction of their fees, websites like Betterment or Wealthfront promised to take a retail investor’s paycheque, allocate it across index funds, rebalance it on autopilot and even optimise for tax. The pitch was simple: stop paying humans to second-guess the market, let software do the boring work for you.
It wasn’t the first time we handed over financial decisions. John Bogle’s index funds in the 1970s had already made stock-picking optional. Nathan Most’s ETFs in the 1990s turned entire indices into single tradable shares. Each step pushed retail investors further away from manual tinkering and closer to automation, and from choosing to delegating.
Fast forward to today, and another handover is unfolding. Only this time, we’re delegating directly to AI agents that live on-chain and not to Wall Street managers or passive wrappers. These AI agents can execute thousands of trades, reallocate funds between lending pools, or rebalance your portfolio after knowing your expectations thoroughly, all this while you sleep.
TOKEN2049 Singapore Is Here!
This October, the global crypto community converges at Marina Bay Sands for the world’s largest crypto event: TOKEN2049 Singapore.
Join 25,000+ attendees, 500+ exhibitors, and 300 speakers as the entire venue transforms into a festival-style pop-up city, featuring a rock-climbing wall, zipline, pickleball courts, live performances and wellness sessions.
We bring to you an exclusive 15% off on the tickets.
The unmatchable speaker lineup includes Eric Trump and Donald Trump Jr. (World Liberty Financial), Tom Lee (Fundstrat CIO), Vlad Tenev (Robinhood Chairman & CEO), Paolo Ardoino (Tether CEO), Balaji Srinivasan (Founder, The Network State), and Arthur Hayes (CIO, Maelstrom), with many more to be announced.
Don’t miss your chance to be part of the defining crypto event of the year.
Use code ‘TOKENDISPATCH15’ for an exclusive discount.
What I see AI agents do on the chain reminds me of my reflection piece on a podcast episode where Mode Network’s James Ross unpacked how AI agents might reshape the way traders interact with DeFi.
Read: From Co-Pilot to Auto-Pilot 🧑✈️
Ross had given an example of Giza’s AI agents, which managed $475 million in volume when I wrote the piece last month. Today, the agentic volume under management has grown more than 2x to $1.47 billion in exactly a month.
Centrifuge DAO, a platform for tokenised real-world assets, in July announced its proposal to pilot a $500,000 USDC allocation with Giza’s ARMA, the flagship stablecoin yield agent. This shows institutions getting more confident in letting AI agents manage real treasury capital in production environments.
All this will happen via non-custodial smart accounts that give the DAO full control while the AI quietly works in the background taking financial decisions on behalf of those organisations so that they can focus on doing business.
There are projects solving similar problems for retail users too.
Almanak is building no-code quant labs where a layperson on the street can spin up hedge-fund-like strategies.
Projects like HeyAnon and Wayfinder are turning DeFi into a text box user interface.
While wrappers like ETFs abstracted away the complexity of stock markets, AI agents go a notch higher and feel like finance brokers — you tell them what you want, and they go execute across the mess of DeFi dashboards, charts, and clicks on your behalf.
So why even bring AI into DeFi?
That’s because DeFi has always felt like a game of whack-a-mole. You chase one yield and another pops up somewhere else. You can sit in front of three dashboards — Aave, Compound, Curve — copy-paste addresses, approve, bridge, wait for the transaction to settle, only to find the rates have already shifted by the time your funds land. Crypto never sleeps, which only makes the challenge exponential. Most of us aren’t built to stay up at 3 AM hunting yields. The ones who make money are either coders running bots or whales who can afford to spread risk across protocols. For the rest of us, DeFi often feels like a maze designed to remind us that we — retail and some smaller institutional investors — are always a step behind.
That’s why projects like Giza and Almanak are seeing adoption.
Giza’s ARMA agent is one of the first times I’ve seen a machine actually do the grunt work in a way that feels tangible. When two protocols on Base got into a “yield war” earlier this year, ARMA moved funds back and forth over 2,400 TIMES IN A SINGLE WEEKEND, eking out an 83% higher return than just parking your stablecoins.
I can’t imagine a human or even a decent bot running that fast without slipping.
Almanak, meanwhile, is attacking a common retail pain point: complexity. Even if you know what you want to do in DeFi, coding up a strategy is a different story. I’ve often stared at quant scripts and thought, “This is not my playground.” Almanak’s pitch is seductive to an average retail investor. Let AI agents design, backtest, and deploy strategies as on-chain vaults, while you just decide the broad strokes. It’s the kind of product that makes me wonder if hedge-fund-level strategies will one day be as easy to spin up as a Spotify playlist.
Today, Almanak’s AI agents manage $35 million in value.
Read: Vibecoding DeFi strategies with Almanak's AI agents 👾
And then there are the chat-first experiments like HeyAnon whose promise is even simpler: talk to DeFi like you’d talk to a friend. Imagine typing, “Swap 100 USDC to ETH and lend it,” and watching an agent do the rest across approvals, swaps, and deposits.
I’ve tried moving funds across three chains before and I can tell you that figuring out approvals, bridges, fees is the kind of experience that makes you want to throw away your laptop.
Wayfinder’s AI agents aren’t stopping at swaps. They’ve rolled out a Perps Agent that lets you trade perpetual futures in plain English: “Go long,” “Set a stop loss at $3000,” or “What’s my current PnL?” It doesn’t need you to open multi-tab dashboards or have expertise in margin trading required. Their Prediction Agent allows you to take positions on trending events with a single command: ETH’s next price swing, whether Kanye was hacked, or if MrBeast will raise $40 million.
They are turning the entire spectrum of DeFi into something navigable by cutting the need for dashboards, spreadsheets and jargon with the help of just everyday language. That feels inclusive in a way DeFi has always promised but rarely delivered.
What ties all of these together is that they aren’t solutions looking for problems. The problems are painfully obvious: too much noise, too many walls, too little time. If these agents can strip that down to intent that reads like “I want yield, I want exposure, I want this asset working for me”, then they’re solving something fundamental. Of course, the sceptic in me knows early DeFi experiments often look smoother in a demo than in the wild. But Giza’s growth, Almanak’s no-code push, and the fact that bigger players are starting to lean in make me believe we’re inching closer to a future where money really does drive itself.
Hype vs Reality
Crypto has a habit of sprinting before it learns to walk. The moment someone utters “AI + DeFi,” the market goes full throttle. Virtuals Protocol is the perfect example. Its token shot up to a $5 billion market cap before reality checked it down to half a billion in two months. Its market cap today hovers around $760 million, still ~85% down from its peak.
For a brief window, its flagship agent AIXBT was the talk of Crypto Twitter. It scraped X (formerly Twitter) and threw in trading signals that people piled into. It did help some people make money during the frenzy, but when the cycle cooled, its market cap also followed the same trajectory as that of Virtuals - 85% down in a couple of months.
I find it hard to forget such experiments. Contrast these with tools like ARMA, which don’t just give signals but quietly execute thousands of transactions and deliver higher yields without anyone needing to babysit. One was an AI tool that provided data to people who used it to gamble, , while the other is like a machine solving the very problem most of us struggle with: time, attention, and efficiency.
Still, I’m not blind to the risks of the latter. Trusting an AI agent with your wallet isn’t trivial. What if it misreads your intent? What if it plugs into a malicious contract and drains everything? For all the talk about permissionless access, DeFi is still unforgiving: one wrong click and your funds are gone. Handing that responsibility to an AI makes the stakes even sharper. I picture a retail investor who already dreads putting any money into crypto after reading all the headlines around wallet drains and exchange hacks.
And then there’s over-automation. If ARMA-style yield hopping becomes the norm, you could see agents competing against each other in the same pool, turning DeFi into a machine-driven knife fight. What’s worse is as retail investors you might not even know that you are in the middle of it.
Regulation is another shadow. If an AI agent is effectively “managing” your portfolio, does that make it an investment adviser? Does the team behind it bear liability if something goes wrong? The US Securities and Exchange Commission (SEC) hasn’t answered those questions, but you can bet they’re coming. We’ve already seen how quickly regulators clamped down on tokenised stocks once they blurred the lines between access and ownership. Agentic finance will attract the same scrutiny.
Despite these challenges, I do think there’s a case for these projects to find a product-market fit.
Every wave of financial innovation at some point started off looking messy, risky, and even unserious. Index funds were dismissed as lazy investing. ETFs were mocked as wrappers for people too clueless to pick stocks. Robo-advisors were seen as bland algorithms that no one would trust with their retirement. Yet, each of those abstractions ended up reshaping the market.
They made participation easier, they lowered the cost of entry, and brought millions of people into investing who would have otherwise stayed out. In the process, they made finance more inclusive and eventually they became invisible infrastructure.
AI agents in DeFi feel like that same turning point. They’re not perfect. Some are untested experiments still learning how to operate safely. Yet the ones that matter are quietly proving that retail investors don’t need to be quants or coders to compete in a 24/7 market. A tool that can flip capital two thousand times in a weekend and deliver a better yield than you’d manage alone can be a gamechanger.
These AI agents can do for DeFi what those older abstractions did for millions of excluded investors. They can take something that feels gated to coders and whales, and open it up for those who just want their money working without the stress of bridges, dashboards, or scripts.
I keep coming back to that idea of self-driving money. Just as most of us don’t care how our Google Maps finds the fastest route, or how autopilot keeps a plane level at 35,000 feet, we won’t care how an AI agent routes our stablecoins across protocols. We’ll just care that it works, that it saves us time, and that it doesn’t break mid-flight.
To me, that’s the bet here. The abstraction layer needs to be solid and complete for finance. If ETFs made markets more accessible and robo-advisors made portfolios more automatic, AI agents might be the ones that make DeFi usable for everyone.
That’s it for this week’s deep dive.
Until then … stay curious,
Prathik
Token Dispatch is a daily crypto newsletter handpicked and crafted with love by human bots. If you want to reach out to 200,000+ subscriber community of the Token Dispatch, you can explore the partnership opportunities with us 🙌
📩 Fill out this form to submit your details and book a meeting with us directly.
Disclaimer: This newsletter contains analysis and opinions of the author. Content is for informational purposes only, not financial advice. Trading crypto involves substantial risk - your capital is at risk. Do your own research.