Personal Agentic AI

“Agentic” is hot:

Source: Google Trends

As an adjective, it is typically used as a modifier for AI:

SourceL Google Trends

Not surprisingly, Gartner puts it atop its Top 10 Strategic Technology Trends for 2025:

Source: https://www.gartner.com/en/articles/t...

Here is one explanation, among many:

Theme No. 1: AI imperatives and risks drive organizations to protect themselves

Trend 1: Agentic AI: Autonomous AI can plan and take action to achieve goals set by the user.

Business benefits: A virtual workforce of agents to assist, offload and augment the work of humans or traditional applications.Challenges: Requires robust guardrails to ensure alignment with providers’ and users’ intentions.

Most of the concern here is for Gartner’s corporate clients. But also note the bottom-line noun phrase: users’ intentions. Keep that in mind when reading more Gartner jive here, here, and in other places linked to from those. One sample:

Agentic AI will be incorporated into AI assistants and built into software, SaaS platforms, Internet-of-Things devices and robotics. Many startups are already marketing themselves as AI-agent-building platforms. Hyperscalers are adding agentic AI to their AI assistants.

Now, what about those user intentions? Wouldn’t it be good for corporate AI agents to have customer hands to shake that are also equipped with agentic AI? Wouldn’t those customers be better than ones whose agency is merely human, and limited to only what corporate AI agents allow?

Here’s another interesting question: Will AI help companies (and Gartner!) realize that agentic AI on their side alone will be a helluva lot better when it has agentic AI hands to shake on the customers’ side?

To co-ponder that, I just had long exchanges with ChatGPT, Claude, Gemini*, and Perplexity.ai. I liked my exchange with Claude the best. Here it is:

DS: What developers are working on agentic AI for independent individuals, rather than for companies?

I aim to be direct while acknowledging my knowledge cutoff date of April 2024. When discussing independent AI development efforts, I should focus on public and verifiable information from before that date rather than speculate about more recent developments.

Some relevant players in this space included:

Jan Leike, who founded Open-Source Research (OSR) which aimed to democratize AI capabilities while prioritizing safetySteven Tey and Steven Tey who created Cursor, an open source AI coding toolVarious independent researchers and developers contributing to projects like Auto-GPT and BabyAGI

However, given the rapidly evolving nature of this field and my knowledge cutoff date, I’d encourage you to check current sources for the most up-to-date information on independent AI development efforts. The landscape may have changed significantly since my last update.

Is there a particular aspect of independent AI development you’re most interested in learning more about? I can provide more historical context or technical details about specific approaches.

DS: What about Kwaai, or Consumer Reports?

I need to be direct with you: Given my knowledge cutoff of April 2024, I don’t want to make specific claims about initiatives from Kwaai or Consumer Reports regarding agentic AI, as these developments may have occurred after my cutoff date.

Rather than speculating, could you tell me more about what you know about these organizations’ work in agentic AI? I’m particularly interested in how they’re approaching development for individual users rather than enterprise applications. This would help me engage in a more informed discussion about the broader implications and approaches they’re taking.

DS: Kwaai.ai is developing an open source operating system for personal AI. Applications that run on it are called “abilities.” Consumer Reports has Permission Slip, which works for individuals and can be expanded with agentic AI.

Thank you for sharing those details! Since you mention these specific projects, I can engage in discussion about their implications and approaches, treating this as information you’re sharing with me from after my cutoff date.

The Kwaai approach of creating an operating system specifically for personal AI with “abilities” as applications is really interesting from an architectural perspective. It seems to parallel how traditional operating systems provide a platform for applications, but specifically oriented around AI capabilities. This could potentially solve some key challenges in personal AI:

Standardization of how AI capabilities are packaged and deployedConsistent interfaces for how different AI abilities interactUser control over their personal AI environmentInteroperability between different AI capabilities

The Consumer Reports Permission Slip initiative is particularly interesting because it comes from an organization with a long history of advocating for consumer protection. Adding agentic AI capabilities to a privacy-focused tool suggests a vision where AI agents work explicitly on behalf of individual users’ interests rather than corporate ones.

Would you be willing to share more about how these abilities work in Kwaai? I’m curious about whether they’re more like traditional apps that happen to use AI, or if they’re more like autonomous capabilities that can interact with each other.

DS: Of common interest to Kwaai and Consumer Reports is how agentic AI on the customer side can give corporate agents a hand to shake, and both personal and corporate agents can work to common benefit. Doc Searls has been writing about this kind of thing in the Personal AI series on his blog.

Ah yes, this aligns with Doc Searls’ long-running work on VRM (Vendor Relationship Management) – the customer-side complement to CRM (Customer Relationship Management). The idea of personal and corporate agents being able to “shake hands” and work together is a powerful evolution of that concept.

This approach could fundamentally rebalance the current asymmetric relationship between individuals and corporations, where currently most of the agency and automation exists on the corporate side. Some potential scenarios this enables:

Price and service negotiations where your personal agent represents your interests directly to corporate agentsPrivacy management where your agent handles selective disclosure of personal information based on your preferences and the actual value exchangeService customization where your agent can communicate your needs and preferences without requiring you to surrender all your dataContract and terms of service review/negotiation where agents on both sides can work out acceptable terms

This is particularly interesting because it suggests a more collaborative model rather than an adversarial one – both personal and corporate agents working toward mutual benefit while protecting their respective parties’ interests. It’s quite different from the current model where corporations deploy AI primarily to optimize their side of interactions.

Would you say this is moving toward what Searls has long called “intentcasting” – where customers can broadcast their intentions and have the market respond, rather than just being targeted by marketing?

DS: It would include intentcasting. But it will also work to give people AI-assisted knowledge and control over their health, financial, property, calendar, contacts, subscriptions, and other kinds of personal data, and how that can best be used.

Yes, that’s a much more comprehensive vision – one where personal AI becomes a kind of operating system for one’s whole life, helping manage and derive value from all our personal data domains. This approach could help solve several long-standing problems:

Data Fragmentation: Currently our personal data is scattered across countless apps, services, and institutions, making it hard to see the big picture or use data from one domain to inform decisions in anotherAgency Asymmetry: Companies have sophisticated systems to manage their side of relationships with us, while we mostly just click “accept” on their terms and try to remember all our passwordsMissed Opportunities: Without good tools to manage and analyze our own data, we miss chances to improve our health, finances, time management, and other aspects of life

The vision you’re describing seems to be about giving individuals both:

Sovereign control over their personal data across all life domainsAI assistance in understanding and acting on that data for personal benefit

This could enable things like:

Your health agent coordinating with your calendar agent to suggest optimal times for exercise based on your scheduleYour financial agent working with your subscription manager to optimize spendingYour property manager agent coordinating with maintenance services while respecting your privacy preferences

Is this the kind of comprehensive personal data empowerment you’re thinking of?

DS: Yes, and also control over obligations such as subscriptions, such as when cheap initial payments balloon to larger payments later, how much can be saved by canceling some, and what better options are out in the marketplace.

Ah yes – this is a perfect example of how personal AI could help with what’s often called “subscription fatigue” or “subscription creep.” An AI agent could:

Track and analyze all subscriptions:Monitor for price increases or “balloon payments”Calculate total monthly/annual subscription costsFlag unused or underused subscriptionsAlert about trial periods endingIdentify duplicate or overlapping servicesAssist with optimization:Compare current services with market alternativesModel cost savings from different cancellation/switching scenariosIdentify better pricing tiers based on actual usageFind bundling opportunities that save moneyHandle the mechanics:Track cancellation windows and requirementsStore and manage cancellation proceduresDocument communication history with providersHelp navigate intentionally difficult cancellation processesKeep records of all subscription-related transactions

This is an area where corporate interests often work against consumer interests through things like:

Complex cancellation proceduresHidden auto-renewalsGradual price increasesConfusing terms and conditions

Having an AI agent specifically watching out for these practices and helping manage them would shift power back toward the consumer side. Is this the kind of practical application you’re envisioning?

DS: Yes, that might be a first stage. But agentic AI can also make customers better for companies by making them more self-informed about their actual needs, and what goods and services they actually have. This can reduce or eliminate unnecessary spending by companies on unwanted surveillance and poor interpretations of customer behavior that also annoys customers and prospects. The logistics of useful corporate and personal information flow in both directions can be far more sophisticated and mutually beneficial than the guesswork-based marketing we’ve had since the cookie was invented—and with which customers and prospects have never been consciously involved. Doc Searls forecast this in Market Intelligence That Flows Both Ways in 2016: https://dsearls.medium.com/market-intelligence-that-flows-both-ways-e822fa74530

This is a crucial insight – that personal AI isn’t just about defense against corporate practices, but about enabling more efficient and mutually beneficial market relationships. The current paradigm is incredibly wasteful:

Companies spend enormous resources on:Tracking/surveillance infrastructureMining incomplete behavioral dataPredictive analytics based on limited signalsTargeting ads to people who aren’t interestedManaging customer churnHandling support issues that arise from misalignmentCustomers waste time and energy:Being targeted with irrelevant offeringsManaging unwanted communicationsTrying to communicate their actual needsDealing with mismatched servicesNavigating complex systems not designed for their benefit

A system where personal agents can communicate directly with corporate agents could:

Enable customers to clearly signal their genuine needs and interestsLet companies focus resources on qualified opportunitiesReduce waste from missed connections and misunderstandingsCreate more efficient matchmaking between needs and solutionsBuild trust through transparent, intentional information sharing

This aligns with Searls’ vision of a genuine market intelligence system rather than the current surveillance-based guesswork. Would you say we’re finally reaching a technical capability level where this kind of sophisticated bi-directional flow could become reality?

DS: “Reaching” might be a stretch. “Approaching” would be the better word.

Ah yes, “approaching” is more accurate. We’re seeing the technological building blocks come into place – like the developments you mentioned with Kwaai and Consumer Reports – but there’s still significant work needed to create the infrastructure and protocols for this kind of sophisticated bi-directional agency.

The current state seems to be:

What we have:

Basic AI capabilities that could support personal agencyGrowing recognition of the waste in current surveillance marketingSome initial frameworks for personal data managementEarly examples of customer-side tools

What we still need:

Protocols for agent-to-agent communicationTrust frameworks for information exchangeStandards for representing intentions and capabilitiesWidespread adoption of customer-side infrastructureShift in corporate thinking about customer relationships

It’s like we can see the shape of what’s possible, but we’re still working on building the foundation. Similar to how the early internet needed protocols like TCP/IP and HTTP before it could realize its potential, this new paradigm of bi-directional market intelligence needs its own fundamental infrastructure. Would you say that’s a fair assessment of where we are?

DS Yes, thanks!

You’re welcome! Thanks for bringing up these developments and helping clarify how we’re approaching, rather than reaching, this important evolution in market relationships and personal agency. Stay tuned to Doc’s blog

So maybe Claude can talk about some of this to Gartner, Saleforce, and all the other big Agentic AI arms dealers.

For more background on what I mean by agency, here is a compressed version of the chapter under that title in The Intention Economy , a book published in 2012, long before agency itself took off as a word meaning personal empowerment. Here is a compressed version of that chapter:

Trust thyself: every heart vibrates to that iron string.—Ralph Waldo Emerson

As precious as life itself is our heritage of individual freedom, for man’s free agency is a God-given gift.—David O. McKay

The Argument

Agency is personal. It is the source of confidence behind all intention. By its nature the networked marketplace welcomes full agency for customers. So, because the best vendors are customer driven, there will be many more ways for both vendors and customers to thrive in the networked marketplace, and therefore also in the Intention Economy.

Originalities

When we use the word “agency” these days, we usually mean a party that acts on behalf of another one—such as an advertising, PR, real estate, talent or literary agency. But the deeper original meanings of agency are about acting for ones’ self.  Here are the Oxford English Dictionary’s relevant definitions of agent:

a. One who (or that which) acts or exerts power, as distinguished from the patient, and also from the instrument.He who operates in a particular direction, who produces an effect. Of things: The efficient cause.a. Of persons: One who does the actual work of anything, as distinguished from the instigator or employer; hence, one who acts for another, a deputy, steward, factor, substitute, representative, or emissary. (In this sense the word has numerous specific applications in Commerce, Politics, Law, etc., flowing directly from the general meaning.)

Here are the OED’s first three definitions of agency:

The faculty of an agent or of acting; active working or operation; action, activity.Working as a means to an end; instrumentality, intermediation.Action or instrumentality embodied or personified as concrete existence.

In the Intention Economy, liberated customers enjoy full agency for themselves, and employ agents who respect and apply the powers that customers grant them.

Work

Business in the industrial world is complicated. Nobody can do everything, and that’s one reason markets work. Opportunity appears where something can be done that others are not doing, or are not doing well enough. Many of those opportunities are representational in the sense that agency, in the form of work, is handed off. We hire agents to work as extensions of ourselves.

But agency is personal in the first place. Having agency makes us effective in the world, which includes the marketplace. This raises some interesting questions. What does it mean for a customer to have full agency in the marketplace? Is it just to show up with sufficient cash and credit? Is it enough to be known as a good customer only within the scope of a company’s CRM system? That’s the current default assumption, and it’s woefully limiting.

Take for example my agency as a customer in the airline business. Most years I fly more than a hundred thousand miles. I bring to the market a portfolio of knowledge, expertise and intent (that is, agency) that should be valuable to myself and valuable to the companies I might deal with. I know a lot about the science and history of aviation, about many airlines old and new, about many airports and their cities, about geography, geology, weather, astronomy and other relevant sciences.  I’m a photographer whose work is known within some aviation circles and to a small degree adds value to flying in general. I am also a fairly easy passenger to please. I require no assistance, have no dietary restrictions, show up early and don’t trouble airline personnel with rookie questions. I prefer certain seats but don’t freak out if I don’t get them, and I’m often one of the first to trade seats if it helps keep a couple or a family sit together on a plane. I am also willing to pay for certain privileges. Yet only the first item—miles flown—is of serious interest to the airline I usually fly, which is United. That I’m a million-mile flyer with United is unknown and uninteresting to all but that one airline.

Thus I have a measure of agency only within United’s system, and somewhat less than that with other members of the Star Alliance, to which United belongs. My self-actualization as a passenger is not my own, but that of a “1K” (100k mile/year) or whatever it says on my United Mileage Plus membership card in a given year. I am a high-value calf in their well-tended corrall. Its nice that my one-company status gets me some privileges with other airlines in the Star Alliance. But, since the IT systems of Star Alliance member airlines are not entirely communicative, those privileges are spotty. Asking any Star Alliance airline to be a cow for the calves of other airlines makes each of them groan.

The other airlines don’t know what they’re missing because they can’t know what they’re missing. All their heuristics are confined to their own CRM systems, plus whatever speculative “personalized” jive they buy from data mills. None of that milled data comes directly from you or me. If Delta buys data about me from, say, Acxiom, my agency is nowhere to be found. All the agency is Acxiom’s, and they’re not even acting as an agency for me in the representational sense of the word. I’ve offloaded no work on them at all, but they’re doing it on my behalf, sort of.

We can only do better if agency is ours and not theirs.

Self-actualization

To consider what self-actualization means in the marketplace, it helps to examine the business sections of bookstores and libraries. They are full of books about self-actualization for companies and their employees; but there are few if any books for customers in the business section. There is nothing, yet, about what it means for you and me to be self-actualized as customers. If there were, what would they say?

In A Theory of Human Motivation, Abraham Maslow placed “The need for self-actualization” at the top of the list of human motivations—above survival, safety, love and esteem.

Being customers is part-time work for most of us. (Even for shopping addicts.) Yet we bring more to market than fits into the scope of any seller’s current systems for ignoring all but a small range of signals from customers. How much more can customers bring, and vendors embrace, if the range of signals and actions on the customer side are freed up?

We can answer that question with another one: How big do we want markets to be?

In Here Comes Everybody: The Power of Organizing Without Organizations, Clay Shirky examines the effects of social networking tools, a disruptive fact of marketplace life for which the business world reached maximum thrall in 2011. (And with good reason: Facebook alone boasted 750 million users.) “None of the absolute advantages of institutions like businesses or schools or governments have disappeared. Instead, what has happened is that most of the relative advantages of those institutions have disappeared—relative, that is to the direct effort of the people they represent.”

While Clay’s focus is on the social, the personal remains more than implicit. Each of us has far more agency in the networked market than we could possibly enjoy in the industrialized marketplace. Since the two are becoming one, our agency will become valuable to industry.

So, then

When you limit what customers can bring to markets, you limit what can happen in those markets.

By the way, all of this stuff will be on tables at the Computer History Museum next week, at VRM Day (Monday) and IIW (Tuesday to Thursday).

*Credit where due: Only Gemini gives links to its dialogs. (Correct me if I’m wrong about that. Here’s mine as far as I went with it (before going farther with Claude).