We’re happy to share the latest in AltsTech’s series profiling how investment managers are using AI, tech, and analytics to generate alpha. We’re fortunate to interview Gelila Bekele, Founder of Aone Partners.
David Teten: Please give us an overview of your firm.
Aone Partners is a search fund partnership backed by 14 enterprise software CEOs, tech founders, and institutional investors in the U.S. Our investors include leaders from Trilogy Search Partners, Pacific Lake Partners, Maven Equity, Datacor, BK Growth and other private investment firms collectively managing over $1 billion in AUM.
With Aone, my mission is to acquire one business and to bring all the resources and expertise of the partnership to invest in growing this business. We are positioned to do both a majority or full equity buy out. We’re best positioned to acquire businesses that fit our criteria: Bootstrapped enterprise software or tech-enabled services businesses
- Less than $1M in financial capital raised
- EBITDA: $1 – $10 million
- Annual Revenue: $3 – 20 million
- Growing and profitable businesses
- High Recurring Revenue (ARR 60%+)
- Diversified Customer base (Top 10 customer: <20% of total revenue)
David Teten: What’s your background? How and why are you in your role today?
Gelila Bekele: I grew up in Addis Ababa and moved to New York in 2012 as a recipient of the Kluge Scholarship, an award granted to 7% of admitted students at Columbia University. After four years, I completed a BA in Economics and Computer Science and joined BlackRock’s Multi Asset Strategies and Solutions Team in New York. For 3 years, I worked on developing capital allocation policies and investment governance frameworks at BlackRock for CIO’s of the world’s largest Sovereign Wealth Funds, Endowments and Foundations.
After BlackRock, I moved to California and completed an MBA at Stanford’s Graduate School of Business. I then accepted an offer to stay another year as a graduate fellow to partner with Stanford faculty in building the school’s entrepreneurship curriculum. Over the year, I worked with 7 faculty members and heads of organizations such as SoftBank, Roblox and Esusu to publish studies on topics ranging from IPO strategy, sovereign LPs and M&A in services industries. These case studies are now used to educate thousands of graduate students in Stanford’s flagship entrepreneurship courses and are published at Stanford Business Publishing and Harvard Business Review, which attracts 7 million unique visitors monthly, and reaches over 18 million professionals across its digital platforms
After completing my time at Stanford, I moved from California back to New York. During that time, I served as a Director at Banyan Software, a holding company that acquires and grows vertical market SaaS businesses. Soon after, with the backing of 14 world-class investors and software CEOs, I started Aone Partners.
I love working on building new ideas and seeing them come to life. When I started Aone Partners in November 2023, I saw a real opportunity to invest in small and mid-sized businesses built on enterprise cloud ecosystems—companies operating within platforms like Ansys, Autodesk, Salesforce, and other B2B SaaS ecosystems. Over the past year and a half, one thing that’s become increasingly evident is that there are hundreds of thousands of founder-led, bootstrapped businesses, often built 10 to 15+ years ago, that are profitable and built on a strong foundation of products and services developed in a pre-AI era. These companies have already earned the trust of their customers, have stable operations, and strong reputations within their niche. What many of them require today is investment of time, capital, and operational support to modernise their internal systems and bring their businesses into this next chapter so they can keep winning.
I’ve worked on setting up M&A systems and establishing investment processes both within a large holding company and, more recently, within Aone Partners to support my own workflow. When I launched Aone in November 2023, I had the good fortune of doing so at a moment when AI products were evolving rapidly—from occasionally imaginative chatbots to capable research tools with meaningful analytical depth.
David Teten: How are you using AI for industry research?
Gelila Bekele: In researching niche sectors, one of the more practical applications of AI has been its ability to parse and structure unstandardised company data at scale. When sourcing from platforms such as ZoomInfo or Grata, one tends to encounter the same challenge: lengthy company descriptions with very little standardised industry classification—no consistent segmentation by sub-industry, customer type, or vertical.
For lists of 50 to 100+ businesses in highly specific markets, GPT and Claude have proven particularly effective. A well-configured GPT model can process these lists, interpret long-form business descriptions, and segment companies into clean, searchable tables—grouped by sub-sector, customer segment, or commercial function. This allows for rapid construction of sub-industry maps and recategorisation of broader markets in a way that’s both structured and useful.
Both GPT and Claude perform well at identifying patterns in product positioning, service models, and customer base—work that would typically require a small team of analysts and a considerable investment of time. Claude handles long-form content especially well and can convert narrative text into exportable tables. That said, for this specific use case, GPT has proven more consistent and reliable.
David Teten: What do you use to map markets and sub industries?
Gelila Bekele: Once a longer list of 50- 100 companies has been properly organised into structured tables—typically categorised by sub-sector, customer segment, and commercial focus—I’ve found it quite helpful to move beyond the spreadsheet and look at the data visually. For this, I use Markmap. It’s a lightweight, interactive mind-mapping tool, and frankly, more effective than anything I’ve seen produced via Excel or the visual outputs available through GPT. It’s particularly useful at the early stages—whether identifying whitespace, clarifying structure, or simply sharing a clean view of how a market breaks down.
The process is straightforward: once the table is finalised, I convert it into Markmap format, which renders an interactive visual in seconds. You can try it here: Markmap + GPT
David Teten: How else do you use AI for industry research?
Gelila Bekele: For initial immersion in a new industry or sub sector, I prefer to listen to reports. There are now tools, notably Google NotebookLM, that convert long-form material into concise, podcast-style briefings—typically 20 to 30 minutes, sometimes shorter. They draw from multiple sources and produce a well-structured audio summary, often voiced as a dialogue. The objective is to get up to speed quickly, and then decide where to go deeper. I was introduced to it by an undergraduate student intern who uses it for academic research. It’s increasingly popular among students, and with good reason.
David Teten: Do you use custom-trained GPTs?
Gelila Bekele: For the queries I run regularly,I have a library of GPTs trained around my standard workflow. They take about four minutes to set up each. Here are few examples:
Preliminary Due Diligence
- This GPT is built to do early-stage deal review. It’s trained on my standard investment criteria, diligence workflow, and scoring logic—and is designed to process high volumes of information and return a structured summary, key flags, and a scorecard based on a custom framework.To protect sensitive information, it’s best to switch off GPT’s data training mode. For good measure, you might also refer to the business using a project/ code name rather than its actual name.
- It works best with additional engagement and prompting. Initial outputs are starting points; with additional prompting, the GPT becomes more aligned to your preferences and sharper in its analysis over time.
- If you’d like to change the scorecard, you can do so by navigating to Edit GPT → Instructions, and scrolling to the embedded table of metrics—edit directly and save a new version. Due Diligence X AI GPT
Advanced prompt engineering
- One of my go-to tools is the Prompt Architect Pro GPT, which draws from 20 advanced prompting techniques to structure better GPT prompts.
- Giving GPT clear instructions is the single best way to improve its performance. Prompt Engineering GPT
Nardwuar GPT for Meetings
- Nardwuar is a Canadian journalist known for disarming artists with unexpectedly specific facts about childhood memories and long-forgotten affiliations. He’s equal parts entertaining and unnervingly meticulous. Inspired by that approach, this Nardwuar GPT is set up to support meeting preparation in the same spirit.
- Paste in anyone’s LinkedIn profile, and it conducts targeted web searches to surface unexpected context, personal connections and also useful personalized gifts.
- You can test it with your own bio. For the best results, copy your entire linkedin profile and paste it into the GPT. Nardwuar GPT
Internal FAQ & Onboarding GPT
- GPTs can be shared across teams and put to use in practical, time-saving ways—like training one on your company’s Slack history, internal FAQs, SOPs, and documentation.
- Instead of new team members asking the same onboarding questions repeatedly, they can simply query a GPT trained on your own materials. It becomes a living, searchable knowledge base—embedded directly into your team chat or onboarding tools.
David Teten: Beyond research and diligence, where else are you using AI day-to-day?
Gelila Bekele: In addition to the GPT libraries, I use both voice mode and the desktop interface in parallel—depending on whether I’m on the move, in meetings, or working through a new project in real time.
Voice Mode on the Go
- While walking through Central Park or in transit between meetings, I use GPT’s voice mode to prepare for upcoming conversations, debrief calls, or work through more complex scenarios. I speak directly into my AirPods, as one might on a call with a colleague, using a custom-trained GPT pre-loaded with the relevant context. It’s a useful way to refine messaging, test thinking, and clarify decisions. Quietly efficient, rather like having a discreet chief of staff on hand.
Transcription Mode at Conferences and Summits
- This has become one of the more practical applications of GPT when I’m on the road. I attend a number of industry conferences and summits each year, and for a time, relied on a pocket notebook to track meeting notes and follow-ups—workable, though not the most efficient system to revisit at day’s end.
- These days, I’ve adopted a different approach. I begin each day at a conference with a new GPT thread. After each meeting, I’ll dictate the key takeaways, follow-up items, and relevant questions while the conversation is still top of mind.
- As the day progresses, I continue adding to the GPT thread via text and by transcribing reflections, next steps, company names, product references. It becomes a kind of running companion throughout the event. Back at the hotel in the evening, I open the same thread on my laptop and run a prompt: “Group all meetings by category and surface all follow-up items.” When managing notes from 20 or more interactions, it saves hours to have this clearly structured.
David Teten: Which current AI products are software CEOs and investors actually using in their businesses?
Gelila Bekele: Over the last year, I’ve compiled a working database of AI-native products that I’ve personally used or that have been recommended by software CEOs, AI engineers, and investors—all of whom have deployed a select number of these products with measurable success in their businesses. This list offers business owners and investors a clear starting point: a shortlist of high-ROI AI products available today, to make it easier to begin deploying AI thoughtfully and effectively across key business functions. See my Playbook for AI Native Standard Operating Procedures.
David Teten: What role can junior team members play in building out AI tools within organizations?
Gelila Bekele: At a recent hackathon hosted by Lovable, I watched students and early-stage founders build working prototypes of products in under an hour. A task that once took a full day now takes sixty minutes, thanks to no-code tooling and AI-native systems. Interns, if given the space, can deliver at that same pace.
One of the more effective ways I’ve scaled AI integration across Aone Partners has been through the structured involvement of interns at both the undergraduate and graduate level. At one point, we recruited a team of eleven undergraduate and graduate level students at a time to intern with Aone Partners, each supporting different parts of the investment workflow: sourcing, research, sub-sector mapping, data enrichment, preliminary diligence, and early-stage modelling.
This model worked best when paired with clear autonomy. Each intern was given space to lead a small 20% project—exploring how a given process might be improved, simplified, or automated. They helped design internal workflows, eliminate repetitive tasks, and test tools against real use cases.
To test new AI tools and systems, we organised the team into analyst pods, each focused on a distinct category. What was particularly helpful was that many of them brought in ideas for AI products they were already using in their academic work, several of which I hadn’t previously come across. One of those, NotebookLM, I now use regularly for condensing long-form material into structured, conversational insight.
As interns learned the mechanics of product, finance, operations, or diligence, they also shared practical fluency in the tooling—how to use GPTs, build lightweight automation, and structure AI-native processes.
David Teten: Thank you so much for lending so much insight into how you’re using AI to build your firm!
