Speculative Fiction: The impact of AI Agents on work

⏳ Let’s take a trip from the future back to the present… and imagine how work as we know it today will radically evolve

2030: What’s better than trying to find the best employee for a role? Creating the perfect one.

In 2030, the majority of the workflow won’t be recruited, but created—meticulously designed AI agents, custom-trained in-house to align perfectly with the company's goals, culture, and nuances. These agents are far more than tools; they are bespoke entities, endowed with specialized skills and "corporate intuition" that make them uniquely suited for a specific task at a specific company.

The hottest new business unit: The In-House AI Agent Training Lab

Central to this operational success is the role of the AI Trainer, who doesn't just onboard AI agents but shapes their "corporate intuition." These trainers ensure that AI agents aren't just efficient but also aligned with AcmeTech's strategic imperatives and culture. Regularly scheduled touch-points ensure that any drift in AI behavior or understanding is corrected, keeping the machine components of the team as aligned with corporate goals and values as their human counterparts.

The ultimate goal for many organizations will be the ability to train their own specialized AI agents in-house. This will allow for a level of customization, alignment, and agility that pre-trained models can't fully provide.

Imagine internal tooling that allows human team members to tweak the "personality" of a Customer Engagement Agent or fine-tune the parameters that a Data Analytics Agent considers critical. Companies could also imbue agents with highly specific business strategies or compliance requirements.

2026: The future of work is hybrid—part human, part AI agent

Of course, the adoption of AI won't be an overnight phenomenon. It will be a staged process, and one of the key stepping stones towards this future is the integration of pre-trained AI agents that come with specialized, environment-specific capabilities. These agents act as plug-and-play units that can quickly adapt to a company's existing workflows and offer immediate value.

Imagine the marketing division at AcmeTech in 2026—composed of two human strategists which are teamed up with eight specialized AI agents, purchased as a subscription for AI team augmentation, each serving a unique role.

Here’s what a day in their life could look like—

  • Data-Informed Strategy: The Data Analytics Agent starts the day by delivering a succinct, real-time market analysis to the human strategists. It's not just numbers and graphs but actionable insights given AcmeTech’s current strategies and goals.

  • Idea Generation and Decision Making: Armed with this data, the human strategists convene a virtual meeting to formulate the day's strategic objectives. The room for creative thinking and unscripted decision-making is solely theirs.

  • Dynamic Task Delegation: Instead of a static to-do list, tasks are dynamically assigned to AI agents through a dashboard, offering room for pivoting as the day evolves. AI agents also suggest tasks based on newly emerging data, awaiting approval or modification from their human counterparts.

  • Milestone-Driven Feedback Loops: AI agents don't just execute; they also propose adjustments and enhancements. For instance, the Content Creation Agent will share draft creatives and make suggestions for message fine-tuning based on real-time sentiment analysis. Humans can approve, modify, or rethink the approach.

  • Strategic Refinement: At each day's close, human strategists analyze not just the results but also the process. Feedback given here shapes both immediate tasks and long-term strategies, with learnings also encoded into AI behavior for future initiatives.

2024: The immediate value proposition—replace your agency, not your team

Take the example of a Head of Marketing considering the use of external agencies for various functions like SEO, content creation, or paid ad strategy.

Instead, they can now opt for a team of pre-trained AI agents designed to collaborate and tackle these exact roles. This not only offers cost advantages (and timelines—what takes humans a month, can be collapsed into minutes through AI agents), but also streamlines communication and integration, as these AI agents are built to work in concert with each other and human team members.

Here’s what a specific team—AI agency interaction could look like—

The process starts with a seamless onboarding where the AI agency integrates with your existing tools—be it Slack for communication, Notion for documentation, or various CRMs to understand your customer base.

Human-like Interactions with Superhuman Efficiency

Once integrated, these AI agents initiate a 'Kick-off Meeting,' mimicking traditional agency interactions but in a virtual environment. They outline deliverables, timeframes, and KPIs in a matter you're accustomed to. But unlike a typical agency, your AI agency schedules weekly 'Check-ins' that don't require your time (unless you want them to) but simply show up as comprehensive updates in your preferred communication channel.

Playback, Edit, and Replay for Unmatched Transparency

What sets this AI agency apart is its transparency—each major deliverable comes with a 'Playback Script,' detailing the exact logical steps the AI took to arrive at its conclusions. Imagine getting an in-depth understanding of your marketing strategy as a podcast episode, which you can listen to during your commute. Not convinced with some parts of the strategy? Edit the script to add more context or specify different parameters and 'Replay' the simulation to get updated outcomes in real-time.

Example script generated with an agent prototype we did this week (note: all of this was programmatically generated by agents after we provided a few docs from Dayslice, defined agent skills, roles, etc. — there’s a long way to go here but this illustrates a rudimentary version of what type of value we can deliver)

Marketer_Strategist and Copy_Writer Discuss Tone, Style, and Objectives

Marketer_Strategist: "Thanks for joining me today. We've got a new feature, 'Scheduling on behalf of guests,' launching on November 1st. Our budget for content creation is around $10,000. The client wants to aim for a tone that's friendly, helpful, and confident, kind of like a big sister helping you navigate the complexities of running a service business. How do you think we could translate that into our written content?"

Copy_Writer: "That's an interesting tone to aim for. It makes me think of using conversational language, maybe using second-person perspective to make the reader feel like we're directly speaking to them. We should also think about story-driven content that showcases 'day-in-the-life' scenarios to connect with our target audience. What do you think?"

Marketer_Strategist: "I like that approach, especially the story-driven aspect. Given our target audience consists mainly of solo or small service providers who often feel overwhelmed, these narratives could resonate well. Let's document these ideas as our preliminary guidelines for tone and style."

Output:

Visual and Brand Guidelines for Dayslice's New Feature "Scheduling on Behalf of Guests"

Colors

Warm and Inviting: Utilize a pastel color palette that embodies friendliness and approachability. Primary Colors: Soft teal, muted coral, and lavender. Secondary Colors: Cream and light gray.

Imagery

Authentic Lifestyle Shots: Choose images that depict service providers in genuine settings, showing them both working and enjoying their day. Example: A yoga instructor setting up her studio in the morning light, a graphic behind her showing her calendar neatly organized thanks to Dayslice. Dynamic Shots: To differentiate ourselves, let's introduce dynamic angles and compositions in our imagery. Example: An over-the-shoulder shot of a small café owner looking at his fully booked schedule on a tablet, showcasing the ease of the "Scheduling on behalf of guests" feature.

Text Overlays

Friendly Fonts: Use fonts that are easy to read but also have a casual, friendly appeal. Font Choices: Open Sans for body text, Raleway for headers.

Icons and Illustrations

Simple but Unique: Opt for custom illustrations that match our color scheme, emphasizing the "helping hand" nature of Dayslice. Example: An illustrated hand placing a booking into a calendar, surrounded by other illustrated elements like a clock, a phone, or a happy customer.

Brand Differentiators

Interactive Elements: Use interactive elements like hover animations on CTAs or parallax scrolling to stand out. Real Customer Testimonials: Include real, short video testimonials from our users, adding a level of trust and authenticity that not all competitors may have. Cultural Relevance: Make sure our content occasionally taps into current cultural discussions that align with our values, showing that we’re more than just a tool; we’re part of the community.

By adopting these visual and brand guidelines, we aim to create a distinct identity for Dayslice, differentiating us from other service platforms while also encapsulating our desired tone and style.

Discussion Between Copy_Writer and Brand_Manager

Copy_Writer: "Alright, so we've got our tone, style, and visual guidelines nailed down. We know the launch date for 'Scheduling on Behalf of Guests' is November 1st. Let's brainstorm some specific content pieces. What are you thinking in terms of visuals?"

Brand_Manager: "I'm considering creating a set of custom illustrations that can be used across multiple platforms. These could be particularly useful for blog headers, social media posts, and any promotional materials. It will be consistent with our pastel color scheme and friendly tone."

Copy_Writer: "Love that. And given our 'big sister' tone, how about a series of blog posts that go beyond just the 'how-to-use-this-feature' type content? We can write pieces that help small service providers tackle common challenges, with our new feature woven into the solution."

Brand_Manager: "That sounds perfect. We could also put together a series of Instagram Stories that follow a day in the life of different types of service providers, showing how they use Dayslice. We can include polls and interactive elements to engage the audience."

Immediate Turnarounds

The time factor is nothing short of revolutionary. What traditionally would take a human agency months to deliver—from market research to drafting a multi-channel marketing strategy—AI agents can produce in minutes, all without sacrificing the quality or depth of the work.

Additional Benefits

  1. 24/7 Availability: Unlike a human agency, AI doesn't sleep. Want a strategy shift over the weekend? It's done before Monday.

  2. Scalability: As your business grows, so does the AI agency's capability to manage increased workloads, without the need for you to hire more resources.

  3. Data Security: All your data remains in-house (and on-prem if you want it.) There’s no risk of leaking sensitive information through an external agency.

  4. Cost-Efficiency: Eliminate the overhead costs of traditional agencies and pay only for the services you require.

  5. Customization: Even though they're pre-trained, these AI agents will be fine-tuned to better align with your specific business needs, giving you a semi-custom solution without the custom price tag.

Credit / Supporting Research

Some research papers that radically altered my view of the future:

  • ❣️ A Valentines Day Party for AI Agents (4/7/23) (academic paper, Medium summary) - This paper talks about creating virtual "people" called generative agents who live in a simulated world called Smallville, almost like a high-tech version of the game The Sims. These agents can talk to each other, remember past interactions, and even plan events like a Valentine's Day party. For example, one agent named Isabella decided to throw a Valentine's Day bash and went around inviting her friends. Her friend Maria, who has a crush on Klaus, helps her decorate, and they both enjoy the party when the day arrives. What makes this study really exciting is that it shows how these virtual beings can act a lot like real people, making decisions and forming relationships all on their own. It's like taking storytelling and problem-solving to a whole new level. While this is a very “cute” implementation of multi-agent simulations, you can also apply this model to B2B use cases.

  • đź’» A simulation of a software team builds a software game in <7 minutes and at the cost of <$1 (7/16/23) (academic paper) — This paper explores how large language models can be integrated into the software development process. The authors introduce ChatDev, a virtual environment that uses language models to facilitate tasks like designing, coding, testing, and documenting software. The paper organizes the development process into four main stages and uses "agents" to handle each stage's tasks. Through this setup, ChatDev can complete the software development cycle in under seven minutes and for less than one dollar. The paper suggests that this approach could offer a new, efficient way to tackle software engineering challenges.

  • 🧠 A test of various techniques to drive autonomous cooperation among agents and provide insight into their “cognitive” processes (academic paper) —This paper which explores the challenges (and tests some potential solutions) of agent collaboration. The authors introduce a new system they call "role-playing," which uses a technique known as inception prompting to guide these chat agents. The idea is to make the agents work together, completing tasks while staying aligned with what a human would want.

    The paper also talks about the problems that come up when trying to make these chat agents cooperate. There are challenges like getting stuck in conversational loops or not clearly defining who does what. (This is good because it means this is quite a bit of complexity in getting this right, and it isn’t a solved or trivial problem.)