Shawn GrubbPortfolio Manager| Thrive ProtocolCincinnati, OH, United States
Context: I am cleaning up a thin portfolio of about ~70ish tech projects ($50K to 100K USD each) and I think I need to scale it to 200 to 700 projects within the next 12 months.
Details: the PMs & projects are all external and PM capabilities are all over the place, and we are only managing time and quality, cost is fixed. Think of very simple binary milestone evaluation: did you hit the timeline with quality? If yes, get paid. If no, thinly manage the exception: if it cannot be quickly corrected, cancel the project.
As designed today, PMs need to submit updates every fortnight, milestones are monthly, and all communications are managed via telegram chats (omg!).
Its all transactional - and I need to clean it up and put AI agents in place - somehow.
Saving Changes...
Sort By:
Luis BrancoCEO| Business Insight, Consultores de Gestão, LdªCarcavelos, Lisboa, Portugal
Shawn Grubb This is a fascinating (and very real) scenario - thank you for sharing it so clearly.
A few thoughts come to mind, both from a portfolio management and a system design perspective:
1. There is a practical upper limit - but it's contextual.
It’s less about the number of projects and more about the governance model, signal quality, and system maturity. With transactional oversight, volatile PM capabilities, and communication via Telegram (!), even 70 projects can strain the system.
Scaling to 700 will require massive simplification, automation, and a shift from project-by-project oversight to pattern-based exception management.
2. Your model already suggests the right strategic lens: “binary milestone + thin exception management”.
This is a form of portfolio flow control, not traditional project oversight.
The goal is to manage signals, not stories.
From that angle, AI agents make sense - but only if the data architecture is trustworthy, consistent, and focused on decision-enabling signals.
3. Before scaling, design the portfolio operating system.
Ask: What minimum viable governance will give you confidence across 700 parallel streams?
Consider:
- A project schema that enforces consistency at intake
- Milestone-based heartbeat reporting (no free text; structured signals)
- A triage bot to classify project health and route exceptions
- AI as signal amplifiers, not magic fixes — garbage in, garbage out
4. Communicate through the system, not around it.
Telegram may be practical, but it's a black hole for portfolio oversight.
If you must use it, build a wrapper layer that extracts signals and maps them to your milestone model.
Otherwise, you’re flying blind.
In sum: Yes, 700 projects is possible — but not with the current setup.
You’ll need to shift from project-centric control to signal-based orchestration, supported by lean governance and AI agents designed for high-volume, low-complexity environments.
This is not classic portfolio management — it’s portfolio-scale systems thinking.
Would love to hear more about how you're redesigning the architecture.
Fascinating challenge!
...
1 reply by Shawn Grubb
Jul 23, 2025 8:42 AM
Shawn Grubb
...
Luis
Love this input and thanks! Here is some of what I gleaned from your response:
- Key factors: standard parameters, clear governance, and the ability to evaluate signal quality
- Love the idea of portfolio flow control vs. standard PPM
- Love the idea of exception vs stories (but marketing wants stories too)
- AI factors: Architecture is trustworthy, consistent, focused on decision enabling signals - love this
MVP for scaling:
- Consistency at intake: aside from legacy, we are close to this.
- Milestone heartbeat: we have structured initial, later milestones are proven on-chain.
- A triage bot for health/exceptions: YES! This is what we need - using internal and external data
- AI as amplifiers, not magic: this troubles me, feels like we don't validate AI input.
- Telegram is a black hole for portfolio oversight: 100% agree, I need the AI bot to provide prompts, evaluate responses (or lack of) and elevate gaps.
Net: this is for sure an interesting architecture to build on - my thought process is to:
1. Validate signal quality: are we getting good performance data?
2. Evaluate standards: how standard are we really (and isolate exceptions)
3. Baseline assessment: how heathy are the current 70?
Based on this, tactically focus on:
1. Fix any performance reporting gaps
2. Kill underperforming projects
3. Ensure new projects have the right standards/reporting expectations
4. Governance parameters are simple and verifiable
Strategically:
1. Identify exception triggers
2. Look for AI agents to automate project coms (milestone dunning & "any help needed" prompts)
3. Look for AI agents to evaluate milestone & performance data
Long term: (If I can get the thin later working)
1. Expand governance to include classic PPM parameters
2. Introduce multi-variant milestones and less structured reporting (flexibility)
"This is not classic portfolio management — it’s portfolio-scale systems thinking"
- Exceptionally well phrased, and thank you so much for your feedback Luis!
The limit on a portfolio size is really driven by the ability to make good decisions from information related to it. Very large organizations could have enterprise portfolios with hundreds if not thousands of components, but it is rare that any one portfolio manager would need to manage anymore than a subset of those.
Given the very narrow focus you have on these, a simple spreadsheet would be a very manageable way of dealing with them so long as there are sufficient qualitative characteristics identified as columns to enable filtering down to meet the needs of individual stakeholders.
Kiron
...
1 reply by Shawn Grubb
Jul 23, 2025 8:53 AM
Shawn Grubb
...
Kiron -
Thanks - your post is a reminder to not over engineer. In the past, I ran a $250M IT portfolio using excel macros sourcing data from MS Project. But, we used the method to kill every single smaller project and limited the PF to about 20-40 most strategic projects. While this project is not that, your principle of keeping it simple I think 100% correct. The challenge is as Luis brought up in the prior post - quality signal, simple parameters, thin governance, and use AI to find exceptions.
Shawn, do you have access to any type of PMIS (project management information system)? And what are you looking to accomplish with the AI agents?
...
1 reply by Shawn Grubb
Jul 23, 2025 9:04 AM
Shawn Grubb
...
Amanda -
omg.. our managing org has very little PM capabilities or understanding. But, all proposals, evaluations, decisions, and milestones are captured and managed on the blockchain - so in essence, the blockchain is the "reporting" PMIS. The individual projects are tech startups - and from what I have seen most are high capacity/capability devs, but they often lack operational rigor. I have not yet met a PMP, but I have seen some of them figure out Asana, Monday, Jira, Notion etc...
I think we will need AI agents to manage basic "are you on track" conversations, evaluate reporting quality, and eventually to make governance decisions. Fun stuff!
Saving Changes...
Shawn GrubbPortfolio Manager| Thrive ProtocolCincinnati, OH, United States
Jul 22, 2025 5:25 PM
Replying to Luis Branco
...
Shawn Grubb This is a fascinating (and very real) scenario - thank you for sharing it so clearly.
A few thoughts come to mind, both from a portfolio management and a system design perspective:
1. There is a practical upper limit - but it's contextual.
It’s less about the number of projects and more about the governance model, signal quality, and system maturity. With transactional oversight, volatile PM capabilities, and communication via Telegram (!), even 70 projects can strain the system.
Scaling to 700 will require massive simplification, automation, and a shift from project-by-project oversight to pattern-based exception management.
2. Your model already suggests the right strategic lens: “binary milestone + thin exception management”.
This is a form of portfolio flow control, not traditional project oversight.
The goal is to manage signals, not stories.
From that angle, AI agents make sense - but only if the data architecture is trustworthy, consistent, and focused on decision-enabling signals.
3. Before scaling, design the portfolio operating system.
Ask: What minimum viable governance will give you confidence across 700 parallel streams?
Consider:
- A project schema that enforces consistency at intake
- Milestone-based heartbeat reporting (no free text; structured signals)
- A triage bot to classify project health and route exceptions
- AI as signal amplifiers, not magic fixes — garbage in, garbage out
4. Communicate through the system, not around it.
Telegram may be practical, but it's a black hole for portfolio oversight.
If you must use it, build a wrapper layer that extracts signals and maps them to your milestone model.
Otherwise, you’re flying blind.
In sum: Yes, 700 projects is possible — but not with the current setup.
You’ll need to shift from project-centric control to signal-based orchestration, supported by lean governance and AI agents designed for high-volume, low-complexity environments.
This is not classic portfolio management — it’s portfolio-scale systems thinking.
Would love to hear more about how you're redesigning the architecture.
Fascinating challenge!
Luis
Love this input and thanks! Here is some of what I gleaned from your response:
- Key factors: standard parameters, clear governance, and the ability to evaluate signal quality
- Love the idea of portfolio flow control vs. standard PPM
- Love the idea of exception vs stories (but marketing wants stories too)
- AI factors: Architecture is trustworthy, consistent, focused on decision enabling signals - love this
MVP for scaling:
- Consistency at intake: aside from legacy, we are close to this.
- Milestone heartbeat: we have structured initial, later milestones are proven on-chain.
- A triage bot for health/exceptions: YES! This is what we need - using internal and external data
- AI as amplifiers, not magic: this troubles me, feels like we don't validate AI input.
- Telegram is a black hole for portfolio oversight: 100% agree, I need the AI bot to provide prompts, evaluate responses (or lack of) and elevate gaps.
Net: this is for sure an interesting architecture to build on - my thought process is to:
1. Validate signal quality: are we getting good performance data?
2. Evaluate standards: how standard are we really (and isolate exceptions)
3. Baseline assessment: how heathy are the current 70?
Based on this, tactically focus on:
1. Fix any performance reporting gaps
2. Kill underperforming projects
3. Ensure new projects have the right standards/reporting expectations
4. Governance parameters are simple and verifiable
Strategically:
1. Identify exception triggers
2. Look for AI agents to automate project coms (milestone dunning & "any help needed" prompts)
3. Look for AI agents to evaluate milestone & performance data
Long term: (If I can get the thin later working)
1. Expand governance to include classic PPM parameters
2. Introduce multi-variant milestones and less structured reporting (flexibility)
"This is not classic portfolio management — it’s portfolio-scale systems thinking"
- Exceptionally well phrased, and thank you so much for your feedback Luis!
...
1 reply by Luis Branco
Jul 23, 2025 9:25 AM
Luis Branco
...
Shawn Grubb, what a fantastic response.
I'm genuinely pleased to see how you're approaching this challenge with strategic clarity and a future-oriented mindset.
A few highlights that stood out to me:
- Your emphasis on signal quality as a prerequisite for any meaningful AI enablement
- Framing triage bots not just as automation tools, but as orchestration mechanisms
- The intentional evolution of your system — from binary + exceptions toward a more classic PPM model with built-in flexibility
You're architecting something that could genuinely redefine how portfolios operate in high-scale, low-touch, distributed-complexity environments.
Bravo.
Saving Changes...
Shawn GrubbPortfolio Manager| Thrive ProtocolCincinnati, OH, United States
Jul 22, 2025 6:58 PM
Replying to Kiron Bondale
...
Shawn -
The limit on a portfolio size is really driven by the ability to make good decisions from information related to it. Very large organizations could have enterprise portfolios with hundreds if not thousands of components, but it is rare that any one portfolio manager would need to manage anymore than a subset of those.
Given the very narrow focus you have on these, a simple spreadsheet would be a very manageable way of dealing with them so long as there are sufficient qualitative characteristics identified as columns to enable filtering down to meet the needs of individual stakeholders.
Kiron
Kiron -
Thanks - your post is a reminder to not over engineer. In the past, I ran a $250M IT portfolio using excel macros sourcing data from MS Project. But, we used the method to kill every single smaller project and limited the PF to about 20-40 most strategic projects. While this project is not that, your principle of keeping it simple I think 100% correct. The challenge is as Luis brought up in the prior post - quality signal, simple parameters, thin governance, and use AI to find exceptions.
Thanks - I love your input!
Saving Changes...
Shawn GrubbPortfolio Manager| Thrive ProtocolCincinnati, OH, United States
Jul 22, 2025 8:47 PM
Replying to Amanda Loewy
...
Shawn, do you have access to any type of PMIS (project management information system)? And what are you looking to accomplish with the AI agents?
Amanda -
omg.. our managing org has very little PM capabilities or understanding. But, all proposals, evaluations, decisions, and milestones are captured and managed on the blockchain - so in essence, the blockchain is the "reporting" PMIS. The individual projects are tech startups - and from what I have seen most are high capacity/capability devs, but they often lack operational rigor. I have not yet met a PMP, but I have seen some of them figure out Asana, Monday, Jira, Notion etc...
I think we will need AI agents to manage basic "are you on track" conversations, evaluate reporting quality, and eventually to make governance decisions. Fun stuff!
Saving Changes...
Luis BrancoCEO| Business Insight, Consultores de Gestão, LdªCarcavelos, Lisboa, Portugal
Jul 23, 2025 8:42 AM
Replying to Shawn Grubb
...
Luis
Love this input and thanks! Here is some of what I gleaned from your response:
- Key factors: standard parameters, clear governance, and the ability to evaluate signal quality
- Love the idea of portfolio flow control vs. standard PPM
- Love the idea of exception vs stories (but marketing wants stories too)
- AI factors: Architecture is trustworthy, consistent, focused on decision enabling signals - love this
MVP for scaling:
- Consistency at intake: aside from legacy, we are close to this.
- Milestone heartbeat: we have structured initial, later milestones are proven on-chain.
- A triage bot for health/exceptions: YES! This is what we need - using internal and external data
- AI as amplifiers, not magic: this troubles me, feels like we don't validate AI input.
- Telegram is a black hole for portfolio oversight: 100% agree, I need the AI bot to provide prompts, evaluate responses (or lack of) and elevate gaps.
Net: this is for sure an interesting architecture to build on - my thought process is to:
1. Validate signal quality: are we getting good performance data?
2. Evaluate standards: how standard are we really (and isolate exceptions)
3. Baseline assessment: how heathy are the current 70?
Based on this, tactically focus on:
1. Fix any performance reporting gaps
2. Kill underperforming projects
3. Ensure new projects have the right standards/reporting expectations
4. Governance parameters are simple and verifiable
Strategically:
1. Identify exception triggers
2. Look for AI agents to automate project coms (milestone dunning & "any help needed" prompts)
3. Look for AI agents to evaluate milestone & performance data
Long term: (If I can get the thin later working)
1. Expand governance to include classic PPM parameters
2. Introduce multi-variant milestones and less structured reporting (flexibility)
"This is not classic portfolio management — it’s portfolio-scale systems thinking"
- Exceptionally well phrased, and thank you so much for your feedback Luis!
Shawn Grubb, what a fantastic response.
I'm genuinely pleased to see how you're approaching this challenge with strategic clarity and a future-oriented mindset.
A few highlights that stood out to me:
- Your emphasis on signal quality as a prerequisite for any meaningful AI enablement
- Framing triage bots not just as automation tools, but as orchestration mechanisms
- The intentional evolution of your system — from binary + exceptions toward a more classic PPM model with built-in flexibility
You're architecting something that could genuinely redefine how portfolios operate in high-scale, low-touch, distributed-complexity environments.
Bravo.
Saving Changes...
Sergio Luis ConteHelping to create solutions for everyone| Worldwide based OrganizationsBuenos Aires, Argentina
the key point here is to decide your portfolio management process. for example, if you are using something traditional is not the same than if you are using lean portfolio management. Saving Changes...
Thomas WalentaGlobal Project Economy ExpertHackenheim, Germany
With an anticipated 700 projects, you may want to consider creating a portfolio structure/hierarchy, reflecting the control needs.
A client had an enterprise portfolio with a target of max 50 core projects, spanning multiple divisions (like HR, Finance, supply chain, marketing) and with durations mostly more than one year. This portfolio was managed by the PMO and reviewed twice a year, using portfolio mgmt SW.
Then, each division had its own portfolio of (mainly) divisional projects, typically around 10.
Lastly, the many small projects (3 months or less) that can be handled by the SME teams, such as change requests or maintenance tasks in IT. Listed in the system but not reported on, many did not even have a PM assigned.
Hi Shawn, that definitely sounds like a thorny problem. I think the structural setup such as creating a hierarchy around divisions and portfolios is a good idea, but is there any opportunity to do some 'managing up' or education of the managing organization? Maybe over time, they will give you a bonafide project management tool, that is built to organize what you're describing and make it easier for everyone. Additionally, I'm not sure if the expectation is for *you* to make the AI agent, but if so, this is a helpful guide I've referred to: https://cdn.openai.com/business-guides-and...-agents.pdf
Let us know how it turns out! I'm sure myself and many other in this thread are rooting for you.
