By Sam in AI Agents — 25 Mar 2026

New Battleground: AI Agents are fighting for your computer

2026 is the year AI agents went mainstream. The year they became tangibly productive for the everyday user. The year when the mind blowing possibilities of AI started to become reality.

Photo by Dan LeFebvre / Unsplash

"Success in creating effective AI could be the biggest event in the history of our civilisation. Or the worst. We just don’t know." - Stephen Hawking

Afternoon All,

Google in 2002, Amazon in 2005, Facebook in 2006, and the iPhone in 2007. Years and products that came along and changed how the world operated...forever. And there was no going back. In the future, 2026 will be another year we can point to and say that is when our work and social lives were irrevocably changed. 2026 is the year AI agents went mainstream. The year they became tangibly productive for the everyday user. The year when the mind blowing possibilities of AI started to become reality.

That's what we discuss today

Today's dots:

The AI agent battle to control your desktop
The hidden profit killer and the AI agent built to solve it
The AI agent coming to fix e-commerce's hidden headache

AI Agents are coming for your desktop

Here's the thing: In the space of two weeks, three of the biggest names in AI each shipped a product built around the same idea: an AI agent that lives on your computer, not in a chat window. When Perplexity, Meta, and Anthropic converge on identical ideas, that's a signal about where the entire industry is heading. It would pay to take notes.

Let's unpack that:

On 11 March, Perplexity announced Personal Computer, an always-on Mac Mini running their AI agent around the clock and connected to local files and apps. Five days later Meta launched "My Computer", their agent for Mac and Windows, priced at $20/month. Also capable of reading and editing local files and launching applications. Then on 23 March, Anthropic shipped computer use and Dispatch for Claude, adding screen control, phone-to-desktop task handoff, and connectors to over 50 services.
The shared concept across all three products is telling: SOTA cloud AI handles the reasoning, while the local machine provides access to files, apps, background execution, and the benefits of controlling everything from your phone.
Everyone reading this has received an email from a senior leader asking for information or a report with an unreasonable deadline. The potential of these AI agents means those requests can in theory be handed off and fulfilled almost instantly.
After the viral success of OpenClaw, the subsequent hiring of it's creator by OpenAI, and Jensen Huang (Nvidia CEO) saying this week: "Every company in the world today needs to have an OpenClaw strategy, an agentic system strategy. This is the new computer." It's undeniable, the companies that shape the way we live and work have all come to the same conclusion.
Autonomous AI agents that do the same work you do, on the same computers, but they don't take holidays, sick days or only work 9-5 are not theoretical concepts anymore.

If you remember nothing else: The question about whether AI agents can work on personal computers and do parts of your job is not relevant anymore. The relevant question is now which one does the job best? The final hurdle is giving these agents memory. The product that remembers what you did last Tuesday, what you're working towards, and how you like things done is not far away. When that product arrives...the potential is limitless.

The hidden profit killer and the AI agent built to solve it

Here's the thing: Most distributors believe they know their margins, they don't. A phenomenon called "phantom margin loss" quietly erodes profitability across manufacturing and distribution. Inefficient fulfilment decisions, manual workarounds, and fragmented systems all distort gross margin reporting. What's the solution? Agentic AI combined with a unified commerce platform. Already deployed in a few pilot programs in Walmart and Target, the early numbers are difficult to ignore.

Let's unpack that:

Phantom margin loss happens when businesses make fulfilment decisions without knowing their true costs. Choosing a high-cost carrier, splitting shipments unnecessarily, or routing deliveries out-of-area all chip away at margin in ways that never show up clearly on a P&L. All purpose ERP systems and manual processes help keep these costs hidden.
Unified commerce platforms address this by consolidating inventory, order management, and pricing into a single source of truth. For example a roofing supplies distributor, could eliminate costly emergency freight charges simply by having real-time visibility across its distribution centres. Moving from reactive inventory decisions to proactive rules-based inventory management.
Agentic AI goes a step further than traditional automation. Unlike conventional machine learning that require explicit instructions and periodic retraining, agentic AI operates autonomously. The system is able to monitor signals, simulate scenarios, and execute decisions across multiple functions without the need for human input.

"Unlike traditional AI, these agents are highly adaptive, goal-driven and capable of taking initiative on their own accord."

The technology can continuously process data about supplier performance records, weather conditions, global current affairs, customer demand, and returns at scale. Consequently giving companies the ability to proactively detect disruptions and reroute shipments or adjust production schedules before problems escalate.
Early adopters are reporting significant results: up to 10% revenue growth, 40% improvement in sales productivity, 30% faster order fulfilment, and 50% reduction in customer support costs. These numbers represent significant structural improvements to how the business operates.
Agentic systems making autonomous decisions at scale require responsible governance more than ever. Clear guardrails aligned with government regulations and company policy.

If you remember nothing else: The competitive gap between distributors who adopt unified commerce and agentic AI and those still patching together ERPs and manual processes is growing quickly. The real opportunity isn't just operational efficiency, that's incremental low hanging fruit. The ability to reimagine business operations and uncover previously hidden profit sources...that's the real prize.

The AI agent coming to fix the hidden headache of e-commerce

Here's the thing: When a product suddenly vanishes from search results or a promo campaign fails to deliver, merchandising teams face a time-consuming detective job. Jumping between dashboards, inventory systems, and ranking configurations to find out what happened and why. Constructor's newly launched Merchant Intelligence Agent (MIA) is designed to completely eliminate that investigation so that merchandisers have the answers right at their fingertips.

Let's unpack that:

MIA sits inside Constructor's existing search and product discovery platform. Currently used by retailers such as Sephora and Under Armour it combines a chat-based interface with in-product prompts that give insights without a user even having to ask.
The agent performs three distinct functions:
- 1) Explains why products are ranking where they are.
- 2) Recommends what actions to take.
- 3) Automates selected tasks such as pre-filling campaign set-up fields.
Gen AI is moving beyond customer service chatbots and directly into core operations. Product discovery, ranking, and campaign management are all areas where small ranking changes can have a direct impact on conversion and revenue.

"If a launch happens and something doesn't look right, a merchandiser often doesn't have the time to investigate — they need answers now." — Eli Finkelshteyn, CEO and Co-founder, Constructor

If you remember nothing else: Merchandising teams have long been expected to wear many hats. Strategist, data analyst, systems detective, all at once. Tools like MIA signal a significant trend. AI can now absorb the diagnostic and admin work so that retail expertise can be directed where it actually adds value.