2
min read
Ayfie explores MCP: Making Enterprise AI accessible to every business
MCP lets AI connect directly to the systems your business already uses instead of forcing you to copy everything into a specialized database, and Ayfie's Indexing Layer makes it 99% cheaper and 10x faster, bringing enterprise-grade AI within reach for the 95% of Norwegian companies that have fewer than 50 employees.
Ayfie just announced MCP-powered solutions launching in the first half of 2026, and the early results have changed what we thought was possible: 99% fewer tokens, 10x faster responses, up to 90% lower costs. The combination of MCP and our Indexing Layer means small and medium businesses no longer need to rebuild their tech stack or hire data engineers to get real value from AI. They just connect the systems they already use.
MCP is how we finally bring enterprise AI to small and medium businesses
Yesterday, Ayfie announced that we're bringing MCP-powered solutions to market in the first half of 2026. And I believe this will fundamentally expand who gets to benefit from enterprise AI.
The results we've achieved so far have blown us away. We're talking about making advanced enterprise AI accessible to small and medium sized businesses, and doing it in a way that's faster, cheaper, and more practical than we've seen using RAG alone.
What is MCP, and why should you care?
Let me paint you a picture. Imagine you're running a 30-person consulting firm. You've got client data in your CRM, invoices in your accounting software, project updates scattered across email and SharePoint, and proposals living in various folders. When a client asks for a status update, you're toggling between five different systems trying to piece together the story.
Now imagine asking your AI assistant:
"What's the status of the Johnson project, including outstanding invoices and next deliverables?"
And getting a complete, accurate answer in seconds, pulled directly from all those systems.
That's MCP.
Model Context Protocol is essentially a standardized way for AI language models to connect directly to your existing software systems. Think of it as a universal translator that lets your AI speak fluently with Tripletex, Visma, SharePoint, and dozens of other business systems you already use.
RAG vs. MCP: two different approaches
Traditional RAG | MCP | |
|---|---|---|
Where your data lives | Copied into a specialized database | Stays where it already is |
Setup required | Pipelines, ongoing synchronization, data engineering | Connect the systems you already use |
Strengths | Handles legacy systems, on-prem, deep archives, strict compliance | Real-time access, fast to deploy, low maintenance |
Best for | Large enterprises with legacy and compliance needs | Small and medium businesses that want value fast |
RAG is great, and so far the best and only option for large enterprises with more legacy and compliance demands. MCP takes a different path. Instead of moving your data, it lets the AI connect directly to where your data already lives. The language model reaches out to your existing systems in real time, retrieves what it needs, and gives you an answer.
For small businesses, this is transformative. You don't need a data engineering team. You don't need to rebuild your tech stack. You just connect the systems you already use.
The token problem (and how we solved it)
I need to get a bit technical for a moment, because this is where things get really interesting.
When we started testing MCP, we ran into a significant bottleneck. Standard API implementations were returning massive amounts of data, sometimes up to 500,000 tokens per request. Tokens are the text segments that AI models use to process information, and they're how most AI services charge you.
Processing that much data was expensive, slow, and created too many unnecessary queries.
Think about it this way: if you ask "What did we invoice Johnson Corp last month?" you don't need the AI to read through every invoice you've ever created. You need it to find the specific, relevant information.
That's where the Ayfie Indexing Layer comes in, inspired by our 15 years of experience in RAG (Retrieval-Augmented Generation).
We've developed technology that sits between the MCP connection and your systems. It intelligently filters and organizes the data before it reaches the language model. Instead of 500,000 tokens, we're delivering around 5,000 tokens per request.
The results
Metric | Improvement |
|---|---|
Token usage | 99% reduction |
Response time | 10x faster |
Cost | Up to 90% lower |
This isn't just an incremental improvement. It's the difference between a solution that's theoretically possible and one that's actually practical for everyday business use.
What we're building toward
Our planned launch in the first half of 2026 will open a door that's been closed to too many businesses.
With MCP and the Ayfie Indexing Layer, Ayfie can now offer a cost-efficient and quick-to-implement solution that makes advanced AI accessible even to businesses with fewer than 50 employees. That's a segment representing over 95% of Norwegian limited companies.
Between now and our official launch, we'll continue to refine the technology, expand our integrations, and work with our partners to make sure we're building something that truly solves real problems.
Have questions or ideas about how this technology could help your business? Reach out.
