What is Olli Memory?
Olli Memory is a continuous learning system that watches how you interact with Olli and automatically captures your preferences. You spend less time re-explaining your style and Olli gets more useful with every conversation.
Instead of requiring you to configure rules upfront, Olli infers what you like from the corrections you make and the feedback you give. These are stored as personal learnings that Olli applies the next time a similar output is needed:
Memory works at two levels:
Scope | What it covers | Example |
Personal | How you personally like to communicate and work | "Always frame business cases around cost savings, not velocity" |
Organization | Institutional knowledge shared across your team | "Our competitive positioning emphasizes enterprise-grade security" |
How Olli Learns
Automatic capture from conversation
When you correct Olli or express a preference during a conversation, Olli picks up on it and saves it as a learning. If he doesn't pick up on it, ask him something like 'can you apply that to all similar scenarios in the future?'
Examples of things Olli will capture:
Corrections - "No, don't mention pricing in the first email" or "Frame it around ROI, not speed"
Style preferences - "Keep emails under 3 paragraphs" or "Use a more formal tone"
Terminology - "We call it 'engagement score', not 'health score'" or "Always capitalize Product Platform"
Content preferences - "Always include a case study reference in proposals" or "Lead with the problem statement"
What happens when Olli captures a learning
Olli detects a correction or preference in your message
A capture indicator appears inline in Olli's response, first showing "Capturing preferences..." while processing, then "Captured preference(s)" once saved
A "Manage" link appears on the capture indicator, taking you to the learnings management page where you can review, edit scope, or deactivate the learning
*To change or manage learnings (or 'memories') Olli captures, head to your Settings in Fluint, then 'Olli Memory'.
Reinforcement
The percent bar score is a confidence indictor. If you repeat a preference Olli has already learned, it reinforces that learning, boosting its confidence score. This helps Olli prioritize your strongest preferences when loading them into conversations.
Managing Your Learnings
Accessing Olli Memories
Go to Settings > Olli Memory to see everything Olli has learned. The page has two tabs:
My Learnings: your personal and deal-specific learnings
Organization: org-level learnings (admin only)
What you'll see for each learning
Directive: the specific preference Olli follows
Confidence score: a colored bar showing how strongly Olli weights this preference (green = high, yellow = medium, red = low)
Scope badge: Personal, Deal, or Organization
Type badge: Correction, Style, Terminology, or Preference
Active/inactive toggle: switch to enable or disable the learning
A note at the top indicates that Olli loads up to 20 active learnings per conversation, sorted by confidence. If you have more than 20, lower-confidence learnings may not be active in every conversation.
Editing and managing learnings
From the management page you can:
Toggle active/inactive: temporarily disable a learning without deleting it
Edit the directive text: click the pencil icon to refine what Olli captured
Boost confidence: click the thumbs-up icon to manually increase a learning's confidence
Teach Olli manually
You don't have to wait for Olli to pick up on a preference during conversation. There are two ways to teach Olli directly:
1) From the management page: Click the "Teach Olli" button in the header.
2) From the chat sidebar: Click the brain icon in the sidebar to open the Teach Olli dialog.
Click Save and the learning takes effect in your next conversation.
"What Olli Knows" Panel
While chatting with Olli, click the brain icon in the top-right of the chat sidebar to open the "What Olli Knows" panel. This shows your active learnings grouped by type:
Corrections
Style
Terminology
Preferences
Each section shows up to 3 learnings. If you have more than 5 total, a "+N more" badge appears. From this panel you can:
Click the + button to teach Olli a new preference
Click the settings icon to go to the full management page
Org-Level Learnings
How org learnings are created
When 3 or more people on your team independently teach Olli the same thing, the system automatically promotes that preference to an org-level learning. This runs daily as an automatic process. New team members benefit from institutional knowledge on day one without anyone having to configure anything.
Admin management
Admins and managers can access org-level learnings from the Organization tab on the management page to view and manage shared team preferences.
Tips for Best Results
Be specific: "Use bullet points instead of paragraphs in follow-up emails" is better than "make it different"
Correct naturally: Just tell Olli what you want changed. You don't need special phrasing for Olli to capture it
Review captures: When Olli captures a preference, click "Manage" to verify it's accurate and adjust scope if needed
Give feedback on attributions: When Olli applies a learned preference, use the thumbs up/down on the attribution chip to keep learnings accurate
Curate periodically: Visit the management page to review your learnings. Deactivate ones that no longer apply
Use "Teach Olli": If you have a strong preference, add it directly instead of waiting for a natural conversation moment
Boost important learnings: Use the thumbs-up on the management page to increase confidence on your most important preferences, ensuring they're always loaded into conversations
Known Limitations
Capture behavior
Context sensitivity - Olli may occasionally capture a one-time instruction as a lasting preference. Go to the management page to deactivate it.
Learning cap - Olli loads up to 20 of your top preferences (ranked by confidence) into each conversation. If you have more than 20 active learnings, lower-confidence ones may not be included.
Duplicate detection - Olli checks for near-duplicate learnings before creating new ones, but very differently-worded preferences that mean the same thing may occasionally be stored separately.
Troubleshooting
Olli isn't capturing my corrections
Try being more explicit: "From now on, always..." or "Don't ever..." signals a lasting preference more clearly
Check the Olli Memories page. Olli may have already captured a similar learning previously
A learning was captured that I didn't want
Click "Manage" on the capture indicator to go to the management page
Find the learning and toggle it off or set its status to inactive
Olli isn't applying a learning I've set
Check that the learning is set to active on the management page
If you have many learnings, lower-confidence ones may not be loaded. Boost the learning's confidence from the management page using the thumbs-up icon
I want to share a learning with my team
If enough team members (3+) independently teach Olli the same preference, it will automatically be promoted to an org-level learning via the daily promotion process
I'm a new team member, will I benefit from existing learnings?
Yes. Org-level learnings apply to everyone on the team automatically. You'll benefit from institutional knowledge that your team has built up, starting from your first conversation with Olli.