- Elliot from Ghost Team
- Posts
- 2026 AI Opportunities: 3 Things to Master This Year
2026 AI Opportunities: 3 Things to Master This Year
The playbook 10x-ing your productivity - Skills, MCPs, and ChatGPT Apps
Hey builders,
Wow. A lot can change in one year.
6 months ago if you told me we'd be where we are today, I wouldn't have believed you.
At the end of 2024, it was clear AI would change everything. After our last agency, we saw the opportunity - and knew we had to build something at the centre of it.
By April 2025, I had three thesis:
→ agents and automation will fundamentally change how every company works
→ MCP is critical infrastructure for giving AI access to your data, tools, and systems
→ your business is nothing without strong distribution and personal brand should be at the centre of it
What happened next was a blur. Experimented with a TON of ideas. Launched Ghost Team. Worked with 30+ companies building 80+ enterprise-grade agents and automations.
Doubled down on content and our channels.
→ LinkedIn from 500 → 19.5k followers
→ X from 500 → 2.3k
→ Newsletter from 0 → 3.8k subscribers
→ Launched a new YouTube channel
And brought in 3x the revenue we expected in year 1.
But beyond all that, I've connected with thousands of new people this year
Some of you I've had the pleasure to chat to, many of you I haven't
Some of you are in the Ghost Team community building alongside us every day
But regardless you've been in my corner since the Agent, Automation & MCP space kicked off
All comments, support, letting me land in your inboxes every week
I don't take it lightly, so thank you!
As we kick off 2026, I want to share three things in the AI space that I think are going to be huge this year. If you can master these and get ahead on them, it'll be a big year.
It’s a little bit of a longer newsletter today as I thought these topics deserved some detail.
Let's get into it.
The Building Blocks: Projects, Skills, MCPs, and Agents
Before we get into the picks, I want to level set as a lot of new things have come about this year. We are moving past simple chatbots. To build an AI that actually works for you, you need to assemble these four blocks: Context, Instructions, Access, and Autonomy.

Projects, Skills, App & MCPs, Agents
Project - Who the AI "is"
A dedicated workspace where the AI knows your context, remembers past chats, and follows your rules. "Like hiring an employee who already knows your company, your preferences, and your past work." e.g. ChatGPT Custom GPTs or Claude Projects
Skill - How the AI does a task
Step-by-step instructions for a specific task. Only loads when needed, keeping things fast. "Like a recipe card. The AI follows your exact steps to get consistent results every time." e.g. Claude Skills
App / MCP - What the AI can access
Connections to external tools and services. Lets the AI actually do things in the real world. The top Apps & MCPs can drastically improve what’s possible with AI and leads to a far better experience. "Like giving someone keys to your tools. Now they can book, search, send, and create for you."
Agent - Who executes the work
Specialised AI workers that take on specific roles and run tasks autonomously. You can deploy multiple agents in parallel, each with their own expertise. Agents can be instructed to leverage certain Skills and leverage certain MCPs to get work done. "Like having a team of specialists - a researcher, a writer, an analyst - each with their own playbooks and tool access, working together on your behalf."
Once you understand how these four work together, everything changes.
Projects set the context. Skills define the playbooks. MCPs provide the access. Agents do the work.
Now, let’s get into the most important ones.
My Top 3 Picks to Get Ahead in 2026
No. #1: Claude Skills
Skills are where most teams should start.
We've helped a number of companies set up Projects and Skills this year. It's one of the biggest level-ups we've seen to help teams 10x their productivity.
Full demo on what they are and how to create them here:
Why are they a game-changer? It means no more copying the same instructions and context into every conversation. No more switching between projects. Build once, use everywhere.
Oh and by the way, because Claude has such good models, these skills can do REALLY powerful things.
Here are some examples. Skills can handle basically any repetitive task that needs consistent context:
→ Brand voice refiner for all your content
→ Meeting transcripts into action items
→ Customer pain point extractor from feedback
→ Financial report auto-generator in your exact format
→ Code review following your team's standards
→ Sales proposals matching your template
→ Case study creator from messy inputs
→ Content repurposing across platforms
But here's where it gets crazy - you can stack multiple Skills in one conversation. Here’s an example of how I use them:
→ Create LinkedIn post (Skill #1 loads)
→ Convert to X thread (Skill #2 loads)
→ Generate accompanying image (Skill #3 loads)
→ Crop for LinkedIn's ideal dimensions (Skill #4 loads)
All in one window. All with full context. All working together.
Projects and Custom GPTs lock your knowledge in separate boxes. Skills make your knowledge portable and combinable across every Claude product - web app, API, Claude Code.
Soon, ChatGPT will have skills as well which is great news for anyone or any company who heavily uses ChatGPT.
Start simple. Pick ONE task you do 3+ times per week. Build it, test it, then stack the next.
No. #2: MCPs (Model Context Protocol)
Wow, what a year for MCP.
In just 12 months, it went from an internal Anthropic hackathon project to the primary open standard for connecting AI to external tools - adopted by OpenAI, Google, Microsoft, and hundreds of others. It got so big that Anthropic donated it to the Linux Foundation under a new Agentic AI Foundation to ensure neutral stewardship going forward. That's how important this infrastructure has become.
So, what is it?
MCP is an open protocol that lets AI connect to external tools and services through a standardised interface."
More simply put, its how you plug AI into the tools you already use - your CRM, your analytics, your databases - so it can actually do things, not just talk about them.

MCPs explained
And they’ve been really widely adopted. You can now, directly within ChatGPT, Claude, Cursor, have it access any tool that has an MCP set up:
→ Firecrawl
→ Slack
→ Salesforce
→ HubSpot
→ Perplexity
→ Google Analytics
→ Search Console
→ And hundreds more...
Unlike automation tools like Zapier where "if trigger happens, then action happens," MCPs let the AI decide which tools to use when you chat with it.
For example, if you send a message to Claude with MCPs set up, it decides which of the 200-300 available tools it needs to complete your request. It can execute multi-step workflows, navigate different platforms, and handle complex tasks without you having to spell out every single step.
Think of it like giving the AI the hands and eyes to actually interact with the web, not just talk about it.
Here's a practical demo of me using different MCPs to automate the entire SEO process:
I'd suggest getting started with a setup like this so you can start to see the power of bringing external data into AI.
No. #3: ChatGPT Apps
In October, OpenAI launched their Apps SDK. In December, they launched the ChatGPT App Store.
This is a big deal.
When Apple launched the App Store in 2009, there were 6 million iPhones in distribution. Today, ChatGPT has +850 million weekly users. That's over 10% of the global population.
The opportunity to build inside ChatGPT is enormous, and companies are already rushing to do it.
And this is only the beginning. More of our digital activity is moving inside LLMs - searching, creating, planning, working. Building apps inside ChatGPT, or having your brand present there, isn't a nice-to-have anymore. It's where your users are heading - and it's quickly becoming a customer expectation. The apps that are present will be the ones that get used, capturing an increasing share of attention over time.
OpenAI knows this. They aren't trying to be a better chatbot anymore - they're trying to become the operating system. Just last month they appointed an App Platform lead with one mission: "Turn ChatGPT into an OS."
This is a really exciting space because it's such an early ecosystem. These apps are going to add a huge amount of utility to the way we work within ChatGPT - and later, when they expand to other LLMs and AI platforms as well.
But beyond the utility, there's a big opportunity here. For brands, it's a chance to gain new distribution in a channel that's wide open. For entrepreneurs, it's a chance to be first into an emerging app store before it gets saturated.
The companies that understand this ecosystem first will have a 6-12 month advantage.
We've got a number of resources that can help you understand more about the opportunity and how to build them. Check out appsonchatgpt.com - we launched it as the #1 resource for ChatGPT Apps and the fastest growing community in the space (join the community here).
If you want to build your first ChatGPT App or even prototype what your ChatGPT App could look like, Fractal is a great tool to use. It’s like Lovable for ChatGPT Apps.
We spoke to the team and manage to get a discount code for the community: GHOST20
These will be big themes in 2026!
If you can master them, I have no doubt you’ll be able to 10x your productivity and create great opportunities for yourself and your business this year.
What's Next?
We believe in a future where AI becomes the primary operating system - where interfaces live inside LLMs, not around them. And an entire ecosystem of applications needs to be built for this shift.
Ghost Team will be at the centre of it.
We've been building a platform for the ChatGPT ecosystem in stealth and will be sharing more on it soon. But as our close community, I wanted to give you the first sneak peak.
You can join the waitlist below now!
Thank you for being on the journey and wish you all a great 2026!
Elliot
Learn how your business could grow with Apps on ChatGPT
If your company or brand is interested in getting an app inside ChatGPT, we'd love to talk to you. We have limited spots available to work with ambitious teams who want to win distribution inside of this new platform.


