This weeks issue is a reflection about how AI agent business models will work reflecting on the value they will generate, projects they will highlight, and deliverables they will produce. I find myself thinking about them more like capable consultants who will work on my behalf instead of software for me to directly control. With this mental modal shift it opens new opportunities for focusing on charging for results rather than software.

AI Tools Are Becoming Independent Workers

Remember when software came in a box and we had to install it and figure out how to use it? Then came SaaS (Software-as-a-Service), where we subscribed to cloud-based tools instead. Now we're entering a new era: Results-as-a-Service (RaaS).

RaaS Works Like Hiring a Human Expert

Think of it like hiring a super-smart assistant who doesn't just give you access to tools, but actually delivers finished work. Instead of paying for software access, you're paying for actual outcomes – just like you would with a highly skilled human consultant.

These Services Will Transform Many Industries

Here's what RaaS could look like in practice:

• A marketing AI that actively finds and nurtures new sales leads by developing a sales funnel punctuated with months of discussion and outreach with potential customers to help them convert

• An AI that creates and develops entire software applications by self-prompting itself based on your product requirements documentation

• An academic research assistant that writes, edits, and presents complete strategic market reports tuned to your businesses strategic goals

• A personal AI nutritionist that tracks your health data-points and delivers a personally tailored meal and fitness regimen to meet your specific health goals

Designers Will Shape How People Use AI Services

As designers, we're uniquely positioned to shape this transition. We're moving from designing interfaces for tools to designing experiences for outcomes. It's like the difference between designing a kitchen and designing a meal delivery service – the end goal isn't the tools, it's the result.

Journey mapping will become a crucial skill for designing AI agent experiences. Just as we map human service experiences, we need to map out how users interact with AI agents and the steps these agents take to deliver results. This includes:

• Creating the detailed user personas who will work with AI agents

• Breaking down interactions into clear stages from initial request to final delivery

• Identifying the data and input the AI agents will leverage

• Spotting potential friction areas before they become problems

For example, when mapping an AI design agent's service, you might track:

1. How clients first describe their problem statement

2. The back-and-forth refinement process of meta-prompting to arrive at the project brief for the AI agent

3. How progress updates are delivered to the human client

4. The way final deliverables are presented

5. The feedback, collaboration and revision cycle for the work between the human AI agent

The goal is to make working with AI agents feel as natural as working with highly skilled human consultants, while taking advantage of AI's unique capabilities. By mapping these journeys, we can create experiences that deliver clear value and build trust in AI services.

Three Things Must Happen for RaaS to Work

The capabilities of AI agents are rapidly evolving, but they aren’t a reality yet. Never-mind that highly skilled AI agents have some technical hurdles to overcome from a functionality standpoint, there is also a lack in infrastructure to support this future where anyone can scale-up their knowledge work with highly capable AI agents.

1. AI Marketplaces Need to Be Built Smart platforms where people can find and connect with specialized AI agents for their specific needs. Today, we have UpWork where you can quickly hire experts to help with specific business functions, the first AI agent marketplace agents.com has emerged to solve this early need of agent discovery and hiring.

2. New Payment Systems Must Be Created Secure ways to pay for results rather than time or access. The internet is full of paywalls and for AI agents to navigate this space to accomplish their results they will need to pay for time and materials just like a consultant. Crypto wallets like Solana’s wallet that let users rapidly pay for and exchange in range of internet native and traditional currencies provide clues on how these agents might pay for project materials and receive payment from client users.

3. AI Must Keep Getting Smarter AI systems that can learn and adapt to deliver increasingly sophisticated outcomes. We are still a long way from autonomous agents that operate as skilled consultants, but the future is bending in this direction. Foundational models like Claude Sonnett 3.5 and OpenAI’s ChatGPT Pro are capable of driving results like coding an app or webpage from a prompt to production ready code. Granted, it is not perfect, but it is a proof point that results-oriented agents are on the horizon.

It is is up to us to design the experiences, capabilities and appropriate usage of these autonomous AI agents to ensure they operate in-service of human users and societal needs. This area of innovation is fraught with ethical and safety concerns. In future newsletters we will debate:

• What is the future of work for humans who will no longer need to focus on some of the skilled labor that is automated by agents?

• How do we design constraints into the interface between human users and AI agents to prevent agents from accepting dangerous, illegal or immoral tasks from human users?

• How might we design human and AI agent interfaces that promote collaboration rather than outright replacement of skilled human workers?

These provocative questions are about us designing a future of abundance of knowledge that empowers everyone to have a capable team ready to take action on their behalf at a moments notice. The flip-side is we aim to avoid disempowering large populations that are exposed to wholesale job loss due to autonomous agents.

These Areas Will Adopt RaaS First

Watch for these early signs of change:

• Repetitive tasks that are easy to automate but valuable to complete — such as tax preparation

• Premium services currently only available to the wealthy at a high cost — such as scheduling assistant

• Tasks that require coordination between multiple tools or processes instead of one streamlined platform — such as sales and marketing associate

This Week's Principle Is About Agents Driving Business Value

As AI agents become directly tied to business outcomes and revenue generation, they transform from optional tools into premium services that drive core business value. Companies that harness this shift will gain significant competitive advantages in the near future.

What Makes Results Valuable:

• They solve specific problems that businesses or individuals face

• They create measurable impact (revenue generated, time saved, costs reduced)

• They're delivered in a ready-to-use format

• They require minimal oversight or management

• They're consistent and reliable

Understanding Deliverables:

• A deliverable is the tangible output of an AI agent's work

• It could be a software app, a market research report, or a sales funnel

• The key is that it's complete and actionable

• It should match or exceed what a skilled human professional would deliver

• It needs to integrate seamlessly into existing business processes

Why Results Matter More Than Tools:

• Businesses pay for outcomes, not effort

• Time spent learning and managing tools is time not spent on core business

• Direct results reduce operational overhead

• Measurable outcomes justify investment

• Results can be immediately tied to business value

The Business Model Shift:

• Moving from paying for access to paying for outcomes

• Pricing based on value delivered rather than time spent

• Success metrics tied directly to business goals

• Risk sharing between provider and client

• Scalable delivery of consistent results

This principle reshapes how we think about AI services:

• Focus on defining clear, measurable outcomes

• Design experiences around result delivery

• Build trust through consistent quality

• Create value through predictable delivery

• Scale through reproducible processes

This framework reinforces today’s core principle: successful AI agents combine autonomous operation with direct business value creation. Just like with human workers, the most successful agents will be the closest to the business value creation. The upper-right quadrant represents our target - where AI tools transcend basic automation to become essential business drivers.

Scalable Autonomous AI Solutions

• Create core business value through autonomous operation

• Examples like Vercel and V0 automatically generate deployable applications

• Represent the ideal combination of high outcomes and high value delivery

Basic Outcome-Focused Tools

• Operate independently but produce basic results

• Current LLMs like Claude and ChatGPT fall here

• Limited by general-purpose nature of outputs

Traditional Direct Control Software

• Require manual operation and oversight

• Include standard productivity tools like Microsoft Suite

• Lack direct connection to business value creation

Scalable but Unfocused Solutions

• Offer business value but need significant human direction

• CRM systems and similar platforms fit this category

• Scale well but require manual value extraction

This Week’s AI Agent Radar

Hallucinations aren’t just a technical flaw in an AI agents data or code, they are A design problem. We must design agent logic journey maps to validate an approach, make capabilities clear the goal and restate user questions when unclear. These clarifying steps are just as a high EQ person would clarity and build on ideas to ensure alignment on tasks and outcomes. Conversely, a person with low EQ leaps to conclusions and springs to actions compounding misunderstandings so too can basic agents. Complex agents ask, validate, outline and restate to help clarify before wasting time and compute.

Large world models are here! Google and World Labs, Fei-Fei Li’s new startup, have both published demonstrations of these new models capable of generating 3D game-like worlds from prompts or imagery.

Detailed coverage of World Models via The Sequence. These 3D worlds are game changers for training AI agents in logic functions and how to interact in 3D spaces. It is early days, but the use cases and technology to generate high-quality, realistic 3D spaces using LMMs is powerful new AI capability.

Suggestion box

This newsletter is part of our ongoing series exploring the future of specialized AI through a human-centered lens. Subscribe to join a community of designers and developers shaping purposeful AI agents.

Until next time, keep innovating and stay curious!

Jason and Anastasia