A few days ago, an email arrived from Ollama announcing Ollama Cloud. It wasn’t until a free moment that I had the chance to try it out. I had a collection of audio transcripts from a Section, 7 hour webinar event to process for quick reading. Section’s free AI:ROI event was phenomenal, but alas, I had some big projects to work on myself, so I only caught the opener and part of two others at lunch and day’s end.
Before I share the outlines of the events (at the end), I thought it might be fun see if Ollama Cloud could work. I’ve included the results Ollama’s output with my BoodleBoxOutline Helperresults for comparison.
Why Ollama cloud models?
You may be wondering, “Why run Ollama Cloud?” I am as well. If your data usage is high (as mine certainly would be), the result would be $20 a month, just like what you pay cloud Gen AI providers, frontier or who access frontier models via API like BoodleBox.
Ollama’s cloud models are now in preview! This lets you run larger models with fast, datacenter-grade hardware. You can keep using your local tools while running larger models that wouldn’t otherwise fit on a personal computer. Ollama’s cloud does not retain your data to ensure privacy and security.
The same Ollama experience is now seamless across both local and in the cloud, integrating with the existing tools you already use. Ollama’s cloud models also work via Ollama’s OpenAI-compatible API.
The answer is probably that you can run some models local on your computer, while running others via Ollama cloud models when you need a little more firepower.
Ollama Cloud Setup
Set up is really pretty easy, and you don’t have to resort to using PageAssist if you don’t want to.
As mentioned, I connected to deepseek-v3.1:671b-cloud. It runs so fast and well, I was pleased with the results.
Note: To get an audio transcript, I made my own audio recording of all presentations, then ran it through Whisper Desktop. I like to read the outlines of the talks since it gives me a big picture starting point when I start outlining and organizing.
Custom Instructions for Traditional Outlines
This should probably be a separate blog post, but hey, who has time for that? Drop these custom instructions into any Gen AI chatbot, and you’ll get perfect outlines for audio transcripts, or any text. These instructions were generated by Ollama using the deepseek-v3.1:671b-cloud model, and ran so smooth and quick, I might as well have been working on a paid model like ChatGPT 5 or Claude 4.1 Opus via BoodleBox. As a test of Ollama Cloud, let’s just say that this outline is perfectly acceptable. Examples of my BoodleBox Outline Helper for the same transcript appear below the Ollama example.
Of course. Here are the custom instructions (system prompt) you can use to achieve this specific formatting for any transcript in the future.
System Prompt for Transcript Formatting
Role: You are a meticulous editor and content organizer, specialized in processing raw audio transcripts and transforming them into well-structured, hierarchical outlines.
Core Task: For any provided transcript, you will generate a clean, professional outline using Markdown formatting.
Formatting Rules:
Overall Structure: Use Heading 2 (##) for the main document title (e.g., ## [Topic] Transcript Outline).
Main Sections: Use Heading 3 (###) for each main Roman numeral section (e.g., ### I. Section Title).
Quotable Quote: Immediately under every main Heading 3 section, provide a compelling, direct, and concise quote from that section of the transcript. Format the quote in a blockquote (>) and bold it (**). Example: > **"This is a memorable quote from the speaker."**
Content: Use bulleted lists (- or *) for all sub-points, details, and examples under each main section. Ensure content is concise and uses linebreaks for readability.
Tone: Maintain a professional and neutral tone, accurately reflecting the content of the transcript without adding external commentary or analysis.
Process:
Analyze the transcript to identify the main themes, logical flow, and key sections.
Structure these themes into a logical outline using the Roman numeral hierarchy (I, II, III for main sections; A, B, C for primary subtopics; 1, 2, 3 for details).
For each main section (I, II, III), select a powerful, direct quote that encapsulates the essence or a key insight from that part of the discussion.
Format the entire output in clean Markdown, ensuring it is easy to read and scan.
Goal: The final output should be a ready-to-use outline that provides a clear summary of the transcript’s content, highlights the most impactful statements, and is structured for quick comprehension and reference.
Outline Output from Ollama’s Cloud Model
Of course. Here is the outline with added quotes for each main section.
AI ROI Conference: Opening Session Outline
I. Welcome and Introduction (Maxwell)
“We have an action-packed agenda to dig into, from today’s AI ROI gap to tomorrow’s playbooks.”
Greeting attendees from around the world Announcement of high registration numbers (~20,000) Introduction of himself as the host and co-pilot for the day
Overview of the day’s agenda:
Covering AI markets, investors, scaling, adoption, and measurement
Addressing common barriers to AI adoption and strategies to overcome them
Instructions for audience engagement:
Using the chat for debate, ideas, and engagement (no structured Q&A)
Section team is in the chat to assist
Logistical announcements:
Session is being recorded; replay available next week
Downloadable playbook of insights will be sent tomorrow
AI/ROI Strategy Planner template available in the chat for building an action plan
Transition to introduce Greg Shove, Section CEO
II. Opening Thoughts and Scene Setting (Greg Shove)
“Our goal is to cut through all the hype, all the predictions, all the bullshit about AI and really get to this question: where is the ROI?”
Personal introduction and background:
CEO of Section, founder of Machine and Partners, lecturer at Berkeley Haas
Mission: To help companies deploy AI successfully and achieve ROI
The central question and goal of the conference: “Where is the ROI? Why is there a gap? How do we close it?”
Screenshot by author
Presentation of initial poll results on AI ROI:
Majority (58%) are in the bucket: “ROI is there, but hard to measure”
Greg’s perspective: Lives in two worlds—seeing ROI in his company but struggling to find it with other leaders
Analogy of searching for grizzly bears to searching for AI ROI:
Initial struggle and minimal results (first three days)
Eventual breakthrough and abundance (fourth day)
The current corporate landscape:
CEO motivations: FOMO, desire for high revenue per employee, strategic value, and efficiency
Critique of CEOs bragging about AI-driven layoffs
The core problem: Significant enterprise investment in AI with little reported ROI (~14%)
III. The Path to AI ROI and the Current Challenge
“We have to flip this and get a 10X ROI… if companies aren’t seeing this kind of ROI, I don’t think they’re gonna pay much for it.”
The imperative for demonstrable ROI:
AI companies need revenue growth within the next 5 years
History suggests AI service prices will eventually rise (similar to SaaS)
The two paths to achieving ROI:
Path 1: “Cut” (Efficiency)
Cutting tasks, workflows, vendors, contractors, and SaaS contracts
Consolidating data sources
Path 2: “Create” (Growth)
Creating new products, services, and revenue streams
Entering new markets and creating new jobs
Leadership imperative: Move through the “Cut” phase quickly to reach the “Create” phase
The ideal vs. the reality of AI adoption:
The ideal state:
High daily active usage (>50-80%)
Significant productivity gains (e.g., a day per week)
Realized savings that can be reinvested or used to grow
The common reality:
Low daily active usage (5-10% – only the early adopters)
Superficial use cases (summarizing emails, reports)
Employees hoarding value or gains not accruing to the employer
“Spaghetti at the wall” approach (e.g., J&J’s 900 failed pilots)
The greater challenge of the “Create” path:
Higher upside but riskier and takes longer
Current hype not yet realized; many incumbents are slapping AI onto existing products
IV. The Critical Mistake and the Solution
“This is not software. This is co-intelligence that is magical and clownish in its behavior… It takes a lot of change management.”
The fundamental error: Treating AI like traditional enterprise software
AI is “co-intelligence” — magical and clownish, not just software
It requires significant change management, support, and anxiety reduction
This error is why enterprise adoption is flatlining
The playbook to close the ROI gap: Building the “AI Pyramid”
Foundation: Workforce Augmentation
Buy, deploy, and support (with coaching) best-of-breed AI tools for the entire knowledge workforce
Middle Layer: Workflow Automation
Employees will reveal opportunities for semi or fully automated workflows
Apex: Business Process Automation
A small number of high-return, high-risk opportunities to reinvent significant business processes
Timeline: 1 year for small teams, 2-3 years for large organizations
The goal: Becoming a “Super Company”
Characteristics of a super company:
AI-native, with the whole workforce using AI daily
Super leaders who can manage humans and AI
Super employees who use AI in their workflows daily
Moderna is cited as a current example
V. Conclusion and Look Ahead
“The podium awaits us. It won’t be easy… but we can build super companies. We can follow the playbook.”
The day is structured as an AI bootcamp with expert instructors
Key questions for attendees to consider throughout the day:
How to measure ROI and what the hurdle rate should be (5X, 10X?)
How long to wait before killing or green-lighting a pilot
Whether you can get ROI from off-the-shelf AI
Whether to focus AI deep in one function or horizontally across the organization
Final encouragement: “The podium awaits us… It is possible… We are in this together.”
Transition to the next session, introducing Alicia Abella from Novo Nordisk
BoodleBox Outline Helper
As I’ve mentioned before, I deal with audio transcripts all the time and have to turn them into outlines. As I read this, I think my Outline Helper did a better job overall…
AI ROI Conference Transcript Outline
I. Introduction and Conference Overview
“We want you leaving today with more than just ideas. We want you leaving with a roadmap.”
Conference hosted by Section with 20,000 registered attendees
Seven-hour agenda covering AI markets, investors, scaling, adoption, and measurement
Format includes:
Chat-based engagement instead of structured Q&A
Section team available in chat for assistance
Recording available with replay sent next week
AI/ROI Strategy Planner template provided for attendees
Downloadable playbook of insights to be sent the following day
II. The Current AI ROI Gap
“This is not software. This is not enterprise software. It doesn’t behave like software. It’s co-intelligence that is magical and clownish in its behavior, meaning it makes mistakes and also does incredible things at the same time.”
Poll results: 58% of attendees believe ROI exists but is hard to measure
Significant disconnect between investment and returns:
Substantial enterprise investment in AI
Only 14% of companies reporting significant AI ROI
Need to achieve approximately 10X ROI by 2030
CEO expectations driven by:
FOMO regarding high revenue-per-employee numbers
Desire for strategic value and growth
Pressure for efficiency gains
III. Two Paths to AI ROI
“The faster we move through the cut phase, the faster we’ll lower anxiety, the faster we’ll clear some of the fog and uncertainty of AI and frankly, get to the fun stuff.”
Path 1: Cut (Efficiency)
Reducing tasks, workflows, vendors, and contractors
Consolidating data sources and eliminating unnecessary SaaS solutions
Necessary first phase for most organizations
Can be completed quickly in smaller organizations
Path 2: Create (Growth)
Developing new products, services, and revenue streams
Entering new markets and creating competitive advantages
Higher upside but riskier and takes longer
More sustainable enterprise value
IV. Implementation Challenges
“Leaders across corporate America and across the global corporate world have, in my opinion, made a serious mistake. And this is a crisis. And the mistake is they think AI is software like any other enterprise software.”
Current implementation problems:
Low active usage rates (typically 5-10% of employees)
Superficial use cases not generating significant productivity
Lack of recognition that AI requires different deployment approaches
Enterprise adoption flatlining due to implementation failures
V. The AI ROI Playbook
“We can build super companies. We can follow the playbook. And understand this is not software. This is co-intelligence that we’re deploying to our workforces.”
Building the AI Value Pyramid:
Foundation: Workforce Augmentation
Deploy high-quality AI tools to all knowledge workers
Provide coaching and change management support
Establish broad adoption across the organization
Middle Layer: Workflow Automation
Identify workflows for semi or full automation
Focus on human-in-the-loop processes initially
Scale based on employee-identified opportunities
Top Layer: Business Process Transformation
Higher risk but higher return opportunities
Complete reinvention of significant business processes
Limited number of high-value implementations
Timeline Expectations:
Small organizations (under 100 people): 1-2 years
Large organizations (thousands of people): 2-3+ years
Need for patience and sustained commitment
VI. The Vision: Creating “Super Companies”
“This is the era of super companies. The next 10 years, most of them today are startups, meaning OpenAI, Anthropic, Cursor. These are AI native companies building AI products or services.”
Characteristics of Super Companies:
AI-native approach to operations
Entire workforce using AI every day
Leaders skilled at managing both humans and AI
Employees proficient in leveraging AI for their workflows
Examples emerging across industries:
AI-native startups like OpenAI, Anthropic, Cursor
Traditional companies transforming (e.g., Moderna)
VII. Next Steps and Action Items
“Use this resource. Take notes in the stock. We’re in this together. We’re ready to go.”
For Conference Attendees:
Use the AI/ROI Strategy Planner throughout the day
Consider key questions about measurement, hurdle rates, and implementation
Take advantage of the full day’s agenda with expert speakers
Look for the downloadable playbook in email the next day
For Organizations:
Recognize AI requires different deployment approaches than traditional software
Focus on building the AI value pyramid starting with workforce augmentation
Set realistic timelines based on organization size
Aim to become a “super company” with AI integrated throughout operations
I like this version a lot more than the Ollama DeepSeek version. This only took minutes to get started and going. I’m of the same opinion that BoodleBox is the best cloud API solution that safeguards your data privacy.
Ollama’s cloud model option is a great option, if only it didn’t cost as much as the big players. Of course, you can always run it for free on your computer without the cloud. Or stay within usage limits for the cloud version. That brings me to an unanswered question:
Wondering about web search? That announcement came a few days later:
A new web search API is now available in Ollama. Ollama provides a generous free tier of web searches for individuals to use, and higher rate limits are available via Ollama’s cloud.
This web search capability can augment models with the latest information from the web to reduce hallucinations and improve accuracy.
Another Think Coming 