← sources

LLMs will become the intelligence, persistent memory will become the context, and agents will do the work. What happens to SaaS?

source: reddit (reddit:AI_Agents)published 2026-10-28 00:19ingested 2026-10-29 02:36

retained text

I've been thinking about where AI is heading, and I believe we're approaching a fundamental shift in how we use software.

My theory is that the future AI ecosystem will be built around three distinct components:

1. LLMs provide intelligence.

In real life, people have different strengths.

Some are brilliant at mathematics, others at writing, analytical thinking, programming, or creativity.

We hire people based on their abilities.

Why should AI be any different?

We might choose GPT for certain tasks, Claude for others, Gemini for something else, or specialized models when appropriate.

Intelligence becomes something we select according to the job.

2. Persistent memory provides continuity.

Now imagine hiring a brilliant employee who forgets everything about your projects every morning.

You would constantly have to explain the history, previous decisions, outstanding issues, commitments, and what happened yesterday.

No matter how intelligent that employee is, working with them would be incredibly inefficient.

That's why I believe persistent, incremental memory will become one of the most important layers of the AI ecosystem.

And I don't mean simply remembering conversations.

I mean something closer to a reliable colleague who has worked with you for years and remembers every project:

\- What happened and when.

\- What has already been established.

\- Which decisions were made and why.

\- What changed after new information arrived.

\- What's still unresolved.

\- What needs to happen next.

A memory that evolves with your work, rather than starting over with every new conversation.

Ideally, this memory should be independent of whichever LLM or agent you happen to use.

3. Agents execute the work.

Human activities aren't going to magically disappear.

Someone will still need to answer emails, reconcile bank statements with expenses, prepare documents, plan routes, manage customers, follow up on contracts, and handle thousands of other tasks.

What changes is who performs those operations and how.

Instead of opening ten different SaaS applications, learning ten interfaces, and manually transferring information between them, we'll increasingly ask an AI agent to accomplish a task.

The agent uses the appropriate intelligence, retrieves the relevant persistent context, and interacts with the necessary tools, APIs, and services.

Then it updates the shared context with the results.

Which brings me to SaaS.

I don't necessarily believe SaaS will disappear.

But I suspect many SaaS interfaces will become irrelevant.

The underlying services, databases, transactions, and business logic will still be needed.

What may disappear is the need for humans to interact directly with every application.

Software could increasingly become a collection of capabilities consumed by agents rather than destinations visited by users.

So my mental model is pretty simple:

\- LLMs = Intelligence

\- Persistent, incremental memory = Context and continuity

\- Agents = Execution

\- Humans = Goals, oversight, and decisions

The LLM can change. The agent can change. The tools can change.

But the history of your work shouldn't disappear every time you switch between them.

To me, the real future of AI isn't just smarter models or more autonomous agents.

It's intelligent systems that can pick up where work was left off, understand what's changed, and continue without constantly rebuilding the past.

I'm curious what others think:

Are we heading toward a world where general-purpose AI assistants become the main interface, traditional SaaS products become background services, and persistent context becomes an independent infrastructure layer?

Or will the big AI platforms eventually absorb all three layers themselves?

document fetched 2026-10-29 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1xh9iet/llms_will_become_the_intelligence_persistent