Franklin Templeton Digital Assets says the next layer of value in AI products may sit less in the model itself and more in the harness around it — the software layer that gives a model context, tools, memory, and a way to route a user’s request to the most suitable model.
In the article, translated by ChainCatcher, Julian Love, managing partner of Franklin Templeton’s crypto business, argues for a neutral, decentralized, onchain harness and examines what that shift could mean for AI users, developers, and the broader blockchain ecosystem.
The piece centers on personal agents and consumer finance. Its core claim is that while general-purpose personal agents are moving into consumer use, managing money may call for a specialized agent with different incentives, regulatory standing, and operating infrastructure.
Today, users still have to stitch together fragmented financial information on their own. Most people manage money across separate apps and accounts, without a unified view of their finances, and each action still has to be handled manually.
The article sketches a different model for the future: one agent that a user can contact by text, voice, app, or email, and that can also reach out through those same channels. That agent would understand the user’s full financial picture and take action across accounts to help the user reach specific goals.
It notes that personal agents such as Instinct, Muse, and Grok Bot are already good at everyday tasks tied to money, but they were not built specifically to manage money. It also says most of their business models encourage more spending. A dedicated financial agent may be better suited to the role because a general agent often earns more when the user spends more, while a finance-focused agent may need to succeed by helping the user spend less.
The article adds that crypto infrastructure is a natural fit for agents. Onchain agentic finance is already assembling core components that a financial agent would need, including agent wallets, programmable permissions, automated vaults, and machine payments.
Why personal finance still feels fragmented
For most people, managing money means managing a collection of accounts. The article says one person’s financial life may include a checking account at a bank, a 401(k) account through an employer retirement-plan provider, a brokerage account, two or three credit cards, insurance policies from different insurers, and high-yield savings accounts opened elsewhere to capture better rates.
Each account comes with its own login, app, alerts, and recommendations. None of them really talk to one another. A household’s complete financial picture often exists only in the user’s head, and only when that user has enough time to pull everything together.
That burden slows people down, the article says, or causes them to miss things entirely, which can end up costing real money.
It identifies two developments that are starting to change that. The first is the personal agent — an AI assistant that can understand intent expressed in ordinary language, weigh options, and communicate through multiple channels. The second is onchain agentic finance, which is building the components that let software hold funds, move funds, and earn yield on funds autonomously.
One side provides understanding. The other provides execution. On their own, the article says, neither is enough.
From that starting point, the piece argues that the future of personal finance may be a single entry point: one agent that a user can contact through text, voice, app, or email, and that can also initiate contact through those channels. That agent would know the user’s full financial situation, understand the user’s goals, and act across accounts to help achieve them.
It also argues that the winner in that role may not be today’s leading general-purpose personal agent. Instead, it may be a specialized financial agent running on crypto infrastructure, because the business model of a general personal agent often pushes toward consumption.
The article then lays out four questions in sequence: how people manage money now, what personal agents can and cannot do, why a specialized financial agent may be a better fit, and why crypto infrastructure suits agentic finance.
Where the current system breaks down
Even after nearly 20 years of fintech development, many people still manage their finances through manual work: opening apps one by one to check balances, moving money between institutions to pay bills, remembering due dates mentally, and reconciling everything in spreadsheets when time allows.
Each product may work well on its own. The hard part is the work between products.
The article groups the problem into several points:
- No unified view. Net worth, cash flow, and progress toward goals are spread across institutions that do not share data.
- Reactive rather than proactive management. People often discover a fee, a rate change, or a missed opportunity only after it has already cost them.
- Advice is either expensive or too generic. Personalized planning is priced mainly for wealthier users, while everyone else gets broad suggestions.
- Decisions keep getting delayed. If every action requires research, logins, and transfers, many people choose to do nothing.
- Intent does not turn into execution. A thought like “I should save more” often stays a thought because no tool turns it into a plan and carries it out.
Why general personal agents may not be enough
The first technology that could change this, the article says, is the personal agent. It defines these systems as general AI assistants that act on behalf of users. They can answer questions, do research, draft content, organize tasks, and increasingly take action inside a user’s apps.
Still, when applied to personal finance, the article says they are much better at money-adjacent daily tasks than at directly managing money.
It describes 2026 as a breakout year for general personal agents. Users can contact Instinct by text or phone to make bookings, purchases, and subscription cancellations. Meta’s Muse can send emails, book travel, fill out forms, and complete checkout, both in its own app and in WhatsApp. Grok Bot from SpaceX AI gives each agent its own cloud computer so it can log into a user’s existing apps and operate them the way a person would.
These agents are already touching money-related tasks such as payments, subscription cancellations, and bill negotiations. The article says the contest to become the agent people trust with daily life is only beginning, with Meta and its 3.6 billion daily active users, a startup valued at $10 billion, and SpaceXAI all competing for that position.
How far these systems can go in actual money management remains an open question, in the article’s view. It says several factors point to a new class of specialized financial agent — one that feels similar to today’s personal agents in use, but is built around money management from the start.
Incentives matter
The article treats economics as the most important factor. Companies building general agents usually make more money when users spend more, whether through advertising, transaction fees, or merchant commissions.
A personal finance agent, by contrast, may need to succeed by doing the opposite.
As one example, the article says about 98% of Meta’s revenue in the second quarter of 2026 came from advertising. At the September Connect event, Mark Zuckerberg said Muse would remain free for most users, and Meta expected to monetize it over time by taking a small fee from shopping, booking, and other transactions completed through Muse. Meta also announced new partnerships with retailers including Walmart, Best Buy, Gap, Sephora, and Wayfair.
It says Instinct’s founder stated that the product processes about $1 billion in transactions a year, with more than half tied to travel. At the end of September, Instinct began pushing curated product, dining, and travel recommendations through a feature called Instinct Selections. Some early users criticized the feature, saying they had not asked for those recommendations.
Grok Bot currently sells subscriptions through SuperGrok and Cursor. Even so, the article says its parent company’s IPO filing listed advertising as a core monetization channel for the AI business and said the company planned to integrate ads into real-time conversations.
That creates a tension. An agent whose revenue rises with transaction volume may be an excellent shopping concierge, but it may also be a conflicted financial adviser. A personal finance agent often needs to tell a user not to buy something, and its business model has to support that behavior.
Regulation and depth also matter
The article points to two more factors.
First, moving money, recommending investments, and managing debt can involve licensing, suitability requirements, and in some cases fiduciary obligations. A general agent company that wants to do that work would need to build a regulated financial business alongside its consumer technology business.
Second, personal finance rewards depth more than breadth. A general agent needs to handle thousands of everyday tasks well enough. A financial agent has to reconcile every account, estimate tax consequences based on actual holdings, sequence debt repayment, and make sure every number in every calculation is correct.
That, the article argues, requires purpose-built systems for understanding information, verifying it, and executing on it. It is not just a matter of adding a finance skill to a general agent.
What a finance agent can borrow from today’s products
The article does not argue for starting from zero. It says the strongest financial agents will probably borrow what already works in current personal agents: no need to learn a new interface, text and voice as primary channels, proactive outreach, and persistent memory about the user.
The shape of the product is becoming clearer. What remains unsettled is who will earn enough trust to provide it when the subject is money.
The article says the most likely end state is a division of labor. A user’s general personal agent would manage daily life, while a financial agent would manage money, with the two coordinating with each other.
Agent-to-agent communication is already beginning to appear, it says, citing Instinct’s network for coordinating different user agents as one example. In that setup, if a user asks a daily-use agent whether they can afford a trip, the question could be handed off to a financial agent that actually understands the user’s finances.
Under that model, the general agent acts like a concierge, while the specialized financial agent acts like a chief financial officer.
The section’s conclusion is direct: personal agents are good at money-adjacent daily tasks, but their business models, regulatory posture, and broad product scope make them an imperfect fit for direct money management. A similar user experience, built specifically for finance, may be better suited to the job.
Why crypto infrastructure fits agentic finance
The second major development in the article is agentic finance — software that can hold funds, move funds, and earn yield on funds autonomously.
Traditional finance already has products that resemble parts of this, such as robo-advisers, cash sweeps, and recurring bill pay. But those systems are usually confined to a single institution and are designed for human account holders, not software acting on a user’s behalf.
The article says the more meaningful progress is happening onchain for structural reasons.
Crypto infrastructure is open by default. Agents can have their own wallets and interact directly with any protocol through code, without imitating a user login and without waiting for separate systems to integrate.
The infrastructure is also programmable. Limits set by the user do not have to remain soft instructions. A wallet can enforce them directly. If an agent makes a mistake or is manipulated, the transaction can simply fail to execute.
These systems run around the clock. Stablecoins can settle in seconds, and tokenized assets can generate yield 24 hours a day.
They are also verifiable, because every action is recorded on a public ledger that users can inspect for themselves.
The article also acknowledges real risks, including smart contract vulnerabilities and stablecoin depegs. Managing those risks on behalf of users would be part of the agent’s job as well.
Even so, those same properties are why the first building blocks of agentic finance have appeared onchain. The article says most of these applications still serve traders and crypto-native users, but taken together they already form the execution layer a personal finance agent would need.
Its conclusion here is that crypto infrastructure already allows agents to operate broadly under user-verifiable rules, and onchain agentic finance is building the tools required for execution. What is still missing is an agent that can use those capabilities to manage a person’s full financial life.
A single entry point for money management
The article says that if the usage model of personal agents is combined with the capabilities now being built in onchain agentic finance, and then applied to a product designed specifically for money management, the outline becomes clear: one agent that knows the user, can communicate through multiple channels, and can operate across all of the user’s accounts.
In that model, users no longer have to consolidate accounts themselves. They can talk to the same agent through whichever channel is most convenient at the moment — a text while waiting in line at the grocery store, a voice message while driving home, an app session when they want to go deeper, or an email when documents need to be handled.
When something needs attention, the agent would reach out first, whether a bill is coming due, an interest rate has changed, or progress toward a goal has drifted off plan.
The article says the test for this product is simple: can it take a goal expressed in ordinary language and carry it all the way through to a real result?
What has to be built
To deliver that experience, the article says, it is not enough to place a chatbot on top of a bank account. The product has to get three things right at the same time.
Expert and proactive
- It must understand financial products, strategies, and opportunities in depth, not just at a general level.
- It must anticipate needs before the user notices them, such as falling rates, an upcoming bill, or a cheaper insurance option.
- It must turn user intent and goals into concrete financial outcomes.
- It must keep pushing those outcomes forward instead of waiting for the next prompt.
Persistent and connected
- It must share one memory across channels so a conversation that starts over text can continue by voice without repetition.
- It must stay current automatically as transactions post and balances change, without manual updates from the user.
- It must maintain secure, authorized connections to every institution the user relies on.
- It must run in the background and reach out when attention is needed.
Built around trust
- It must explain each recommendation in plain language, including the trade-offs involved.
- High-impact actions should require user approval, while routine actions can run automatically within user-defined limits.
- It must keep a complete record of what it did and why, and let the user review it.
- It must leave control with the user at all times, with permissions that can be changed or revoked whenever needed.
Why now
The article says acceptance for this kind of product is rising. This year, early users began texting agents and watching them complete tasks.
The supporting infrastructure is also in place, it says: models capable of financial reasoning, expanding open-banking connections across financial institutions, and payment infrastructure that agents are already using.
At the same time, the companies best positioned to reach consumers directly — today’s general personal-agent providers — face structural conflicts of interest when money management enters the picture. That leaves room for specialized entrants, though the article says the window may not stay open for long.
Those entrants may choose to build on crypto first. With US stablecoin legislation now formally in effect, the article says onchain infrastructure can support real operations from day one, while traditional accounts can be added over time through open banking.
Its final point is that the first personal finance agent to win user trust, communicate across channels, and actually help people allocate money and reach goals may end up defining the category. The teams to watch, it says, are the ones that can combine the ease of personal agents, onchain execution, and finance-grade trust mechanisms.
