How Founder-led sales can be done right? • Founder at Sellestial • 3rd-time Founder • Don't be a stranger, follow me and reach out

Is one of your main projects for 2026 cleaning up your HubSpot data?  Want to run AI in production?

If you answered yes to anything, keep on; otherwise, stop reading this.

👉 How we are different to tackle data hygiene with your internal team only:

1. Deep Data Cleaning & Enrichment Expertise
We're data cleaning and enrichment specialists who know this problem space and its traps inside out. By cleaning and enriching 100+ HubSpot portals, we've developed a deep understanding of edge cases that first-timers don't anticipate. This is not general HubSpot admin work; it's pattern recognition from seeing hundreds of messy databases.

2. Dedicated Bandwidth + Project Discipline
Data cleaning is important but rarely urgent, which means it gets perpetually deprioritized by internal teams. We provide dedicated focus with clear project boundaries that prevent the "cleaning to infinity" trap or arbitrary abandonment that internal teams tend to fall into.

3. Production-Grade AI Expertise for Data Maturity
We've built specialized AI infrastructure designed for accuracy and reliability. We are regularly deploying 7+ different AI models for different parts of the process. We tested and evaluated over 50 LLM models over the course of the last few years, and have a comprehensive understanding of their capabilities and caveats. 

The ROI for you: Your internal teams and HubSpot admins can focus on what they do best while we handle the extremely complex, time-intensive data cleanup and/or enrichment with the AI resources your team doesn't have access to.


👉 Want to talk about Sellestial?
Shoot me a DM, and we can schedule an introductory call to discuss data.

Klemen Hrovat

HubSpot quietly dropped 21 AI agents in beta.
Not one. Not five. Twenty-one.
I went through every single one, so you don't have to.

—
👉 Social Post Agent — turns calendar events into network-optimized social content.

👉 Company Research Agent — pulls news, website, CRM data, and competitors into one pre-call brief.

👉 Deal Loss Agent — analyzes closed-lost deals to surface loss patterns by segment, rep, and product line.

👉 Call Recap Agent — turns transcripts into CRM notes, follow-up emails, and flags unanswered questions.

👉 Blog Research Agent — topic suggestions, SEO research, and draft generation inside Breeze Studio.

👉 Customer Health Agent — scores account health from calls, emails, tickets, and gives you talking points + a draft email.

👉 Cross-sell/Upsell Agent — finds expansion-ready accounts and builds tailored outreach with case studies.

👉 Sales to Marketing Feedback Agent — turns objections and deal patterns into actionable themes for Marketing.

👉 Customer Handoff Agent — assembles every conversation, deal, and ticket into one brief when accounts change owners.

👉 RFP Agent — auto-fills proposals using approved answers from past RFPs and knowledge base articles.

👉 Personalization Agent — finds highest-impact personalization opportunities and generates segment-targeted variants.

👉 Knowledge Base Agent — scans closed tickets for content gaps and writes new articles from existing rep answers.

👉 Developer Tool Testing Agent — sandbox for testing custom agent tools with step-by-step diagnostics.

👉 Video Clip Search Agent — finds trending topics, matches them to clips in your library using transcript search.

👉 Closing Agent — answers buyer questions 24/7 from your sales materials, hands off to reps when needed.

👉 Shopify Store Performance Agent — 360° store metrics with industry benchmarks and action recommendations.

👉 Brand Assistant — reviews drafts and flags tone drifts, jargon, or compliance misses against your brand guidelines.

👉 Audit Analyzer — surfaces user behavior patterns and risky activity from your HubSpot audit logs.

👉 Internal FAQ Assistant — centralizes scattered knowledge into one conversational interface with cited answers.

👉 Sales Coach Assistant — reviews call transcripts and delivers personalized coaching with actionable next steps.

👉 ICP Assistant — role-plays as your ideal customer and gives persona-aligned feedback on your copy and campaigns.
—

21 agents. All in beta. All built by HubSpot.
Some are already more polished than others. 
Some might change how you work.

Which one have you tried already? Share your experience in the comments

#HubSpot #AI #RevOps #CRM #agents

Talks about HubSpot data + AI daily @Sellestial | HubSpot Community Champion

HubSpot Solution Partner Bingo.
If you've onboarded more than 3 clients, you've hit at least 9 of these.

I won't spoil all the squares.

But if "Our data is pretty clean" doesn't trigger war flashbacks - you haven't been doing this long enough.

Free space was easy to pick.

If you got bingo on the first read - you're a real one.

P.S. Which square hit the hardest? 

#HubSpot #SolutionPartner #RevOps

You can tell how many different HubSpot admins a company has had by how many different variations of 'DO NOT USE' or 'OLD - ignore' you find in property names.

You can't create a task in HubSpot with Breeze Assistant.
But you can create it with Claude.

Did HubSpot solve one of the longest-standing campaign frustrations?

You can now manually associate contacts, deals, and tickets with campaigns.

From record pages, index pages, the campaigns tool, and even workflows.

If you've ever run a trade show, a field event, or a partner dinner and wondered "how do I tie this back to a campaign in HubSpot?" - that's how.

But here's where it gets interesting.

The attribution rules aren't one-size-fits-all.

A few things worth knowing before you start associating everything.

1) Once a contact is associated with a campaign, they stay "influenced."
Removing the association doesn't undo that. So think before you click.

2) Deals linked through a manually added contact only get attributed if they close after the contact was added. Standard attribution rules apply.

3) Deals directly associated with a campaign? Always attributed. Even if already closed.

That last one is a big deal for RevOps. It means you can retroactively tie closed revenue to campaigns that influenced it offline.

And the workflow action "Add to campaign" means you can automate all of this based on your own criteria. 

No more manual busywork at scale.

This isn't a flashy AI feature.
It's plumbing.
The kind of plumbing that makes your attribution model reflect reality.

What's the first campaign you're going to retroactively associate deals with?

p.s. Available for Marketing Hub Pro and Enterprise. Enterprise gets the extra toggle to enable or disable attribution for manually associated records.

#HubSpot #RevOps #MarketingOps

RevOps is 10% strategy, 20% systems, and 70% explaining to someone why they can't just 'add a quick property'.

HubSpot pros, how do you feel after a week-long OOO?


btw, what did I miss? Which are your favourite updates from last week?

Managing 250k+ contacts in HubSpot and you're still removing duplicates manually?

You're not behind. You're just stuck in a loop.

Here's what I see in portals at that scale:
You clean up duplicates on Monday.
By Friday, new ones are already forming.

Someone imports a list from an event.
Someone else syncs records from a second system. 
A form submission creates a contact that already exists, but with a different email.

The duplicates aren't the problem. They're a symptom.

The real problem? There's no system that understands what "clean" means to you after data gets into your HubSpot.

When you're managing two platforms, let's say HubSpot and another database, with a combined 350k+ records, the question isn't "how do I find duplicates faster?"

It's "how do I stop creating them in the first place?"

That means:
- Enriching records so you actually know what you're looking at before merging
- Building automated filters that catch bad data immediately after it enters the CRM 
- Defining match rules that account for edge cases, not just obvious duplicates
- Reviewing and fixing all the various data sources, bringing data into your HubSpot

Deduplication is the last step, not the first.

The companies that get this right don't run quarterly cleanups. They build a system that keeps the database clean between the cleanups.

Are you still running data cleanup as a project, or have you turned it into a process?

#HubSpot #DataQuality #RevOps #CRM

You can edit records in HubSpot directly from Claude.

Create contacts and other records, update properties...

I like how powerful these integrations already are.

But here's what I actually want from AI + CRM.

Earlier today, I developed a skill in Claude that helps me classify a LinkedIn contact against our ICP. 

As part of the skill, I instructed it to check HubSpot to see if the contact record already exists, so I'm not creating duplicates.

It found an existing contact record in HubSpot.
Noticed the job title was outdated.
Suggested updating it.
And we updated it.

No tab switching.
No "let me go check HubSpot real quick."
No context lost.

The job title update wasn't even the task.
It was a side effect of something bigger.

That's the difference between "AI can update your CRM" and "AI works inside your CRM while doing actual work."

I have been seriously working on and playing with the Claude <> HubSpot MCP connection over the past few weeks. 

The value isn't in the individual actions.
It's in what happens when those actions are part of a larger process.

The real unlock is when your CRM stays accurate because the system self-corrects while doing other things.

P.S. Anyone else running HubSpot through MCP? Which AI are you connecting it to - Claude or ChatGPT?

#HubSpot #MCP #RevOps #DataQuality

Your reps aren't bad at updating HubSpot.

They're just human.

Every two weeks, your sales leader needs to review the pipeline. 
Every board meeting, someone scrambles to make deals look presentable.
Every 1:1, a rep explains a deal they haven't touched in three weeks.

The problem isn't discipline. It's that manual deal hygiene costs more than anyone tracks.

One team I know lost 40+ hours in January to update deal descriptions and prepare a BoD review slide.

We built them the following process.

A deal summary agent that runs on a schedule (every two weeks, automatically), and writes structured output directly into HubSpot properties:

→ Status & Next Steps: the latest activity and blockers
→ Deal Description: a clean, readable summary for BD and leadership
→ Deal Name: standardized, consistent, no more "ACME - Follow up v3 FINAL"
→ Opportunity Potential: what the best outcome could be, based on full deal context

No manual entry.
No stale data in your pipeline.
No scrambling before the board deck.

The reps get time back.
The managers get signal, not noise.
The BoD gets clarity without a pre-call cleanup sprint.

After seeing it in production, the customer said: "It even looks better than when a human was doing it."

That's one example of how we help RevOps use AI.

The data was always there.
The system just wasn't reading it.

#AIRops #HubSpot #AI #RevOps

“AIRops”  “AI for RevOps”  “HubSpot AI”  “AI RevOps”  “AI-first”  “AI-enabled”
these and many more will be use during the Profoundly kickoff.

If any of these resonate with you, welcome.

It starts on March 12, 2026 at 11 am EST

A free, expert-led virtual event focused on how AI is changing RevOps.

Topics:
- What is AIRops
- How HubSpot AI is changing day-to-day operations
- How to use HubSpot as your context infrastructure and massively scale through purpose-built solutions and production agents
- What admins, RevOps leaders, partners, and consultants need to learn
- Real examples from practitioners

If you are:
HubSpot admin
RevOps and GTM leader
Partner or consultant
this one is for you. 

P.s. you might see me during the event, don’t miss that opportunity 🤣

"We need 100% complete data in our HubSpot"

No, you don't.

I hear this from customers all the time. Can you help us get this property to 100% completion?

Sure.
Tell me which random number I should write as "number of employees"

If that data isn't publicly available, I can give you 1,000.
I can give you 5,000.
I can give you whatever makes the chart look full.

But full isn't the same as useful and relevant.

Here's what actually happens when you force people to fill out fields where there's no data available.

We had a client with a required dropdown.
Five values.
Important property for their reporting.

They made it mandatory before a deal could move forward.

Sounds smart, right?

We ran the analysis.
80% of selections were the first value on the list.

Not because it was the right answer.
Because it was the fastest click.

Complete data.
Completely meaningless.

The real problem is timing.

You're asking a rep to choose a close-lost reason on Monday morning when they're not mentally ready to reflect on it. So they pick whatever gets them past the gate.

Maybe they would've given you a thoughtful answer Friday afternoon when they'd set aside 30 minutes to clean up their pipeline.
But you didn't let them choose when.

Here's how I think about all this.

Not every property deserves the same completeness standard.

No contact should be assigned to a deal without a job title.
That's reasonable. 

But requiring job titles for every newsletter subscriber?
There's no point.

A contact entering your CRM doesn't need the same data depth as one that's a week from closing.

The goal isn't 100% complete. 

The goal is
the right data
at the right stage
for the right reason.

Sparse but relevant beats complete but useless. Every time.

P.S. Do you have a property in your HubSpot that you suspect is generating useless data right now?

#HubSpot #RevOps #DataQuality #CRM

HubSpot's Data Quality duplicate tool finds the obvious ones.

Exact same email, domain, company name.

But these aren't the duplicates that typically bother you.

I had a call today with a RevOps lead managing 14,000 companies. 
She tried HubSpot's native tool to find duplicates.
It caught some - the easy ones.

The hard ones?

A company listed as .com and .de.
Same name, but one has "GmbH" appended.
Two records that look different but are the same business.

It doesn't know.
And honestly - neither can you, without opening both websites.

That's the gap.

To resolve real duplicates, you need data outside your HubSpot.

Is .de redirecting to .com?
Is the "GmbH" version a subsidiary or the same entity?
Is the LinkedIn company page the same for both?

This is where AI agents change the game.

Not the "click a button and magic happens" kind.
The kind that opens both websites, checks redirects, compares LinkedIn pages, and gives you a confidence score before merging anything.

The data inside HubSpot indicates that there might be a duplicate.
The data outside HubSpot helps you decide if it actually is a duplicate.

P.S. What's your experience with the HubSpot Data Quality tool to help you with deduplication? 
#HubSpot #DataQuality #RevOps #deduplication

Stop forcing your sales team to enter clean data.

They won't. And even if they try, it won't be clean.

We work inside HubSpot portals every day.
We see what happens when admins add validation rules, required fields, and dropdowns to "ensure data completeness".

Here's what actually happens:
Sales reps are mid-conversation with a customer.
They need to log a job title in HubSpot.
The process demands a perfectly formatted entry.
They're not thinking about your reporting needs.
They're thinking about closing the deal.

So they type whatever gets them through the fastest.

Or worse, you make it a dropdown.
And 80% of your team picks the first value on the list.

You get complete data.
Completely useless data.

Here's what we suggest doing instead.

Let users enter messy data. Typos and all.
"Chief Tech Officer"
"CTO."
"C.T.O."
"Chief Technology Offcer"
all fine

Let them type what feels the easiest in the moment.

Then let the system standardize it behind the scenes.

A month ago, it clicked for me why HubSpot's default country/state property is a free text field.

Not a dropdown.
Free text.

Because data comes in from integrations, imports, forms, and manual edits.
Every source in a different format.

US
U.S.
USA
United States
United States of America

All valid. All messy. All cleanable.

The job of RevOps isn't to be the data police.

It's to build infrastructure that turns messy input into trusted output.

Your sales team shouldn't be responsible for typo correctness.

They should be responsible for having conversations that close deals.

You take the burden of standardization.
That's your job.

P.S. What's one property in your HubSpot that is causing you a "standardization headache"?

#HubSpot #RevOps #DataQuality #DataIntegrity

🎙️ New to streaming or looking to level up? Check out StreamYard and get $10 discount! 😍 https://lnkd.in/dAaAJzce

Email throttling has been standard practice in cold outreach for years.
Now it's coming to HubSpot marketing emails.

I've seen what happens when companies migrate to HubSpot and start sending at full speed on day one. 
Bounce rates spike. 
Spam filters trigger. 
Deliverability tanks before the first campaign even finishes.

HubSpot just released a beta called Throttle Marketing Email
and it's a meaningful addition for anyone sending marketing emails via HubSpot.

Here's what it does
When scheduling a marketing email, you can now set a custom send rate under Advanced Settings. 

Instead of blasting your entire list as fast as possible, the system spreads delivery over time at the speed you choose.

Why this matters more than most people think:
- Sending 50,000 emails in 10 minutes looks nothing like normal human behavior to inbox providers. They notice.
- Large volumes hitting a single company domain at once can quietly destroy your reputation with that domain's ESP.
- High bounce rates hurt way more when they arrive in a concentrated burst. Spread them out, and the damage is manageable.

This beta adds another layer of control to that process. 
It's not just about how many emails you send. 
It's about how fast they arrive.
And if they arrive at all.

Deliverability is somewhat a velocity problem.

Available for Marketing Hub Pro and Enterprise.

P.S. Have you tried the beta yet? Did you wish you could set this in the past?

#HubSpot #EmailDeliverability #MarketingOps #RevOps #Marketing

Unified Revenue View part 2

🎙️ New to streaming or looking to level up? Check out StreamYard and get $10 discount! 😍 https://lnkd.in/dAaAJzce

Today marks a day. For the first time, when I opened my laptop, I went to Claude Code and not to Slack, LinkedIn, or Email 🤯

Dharmesh Shah published his poll results asking his newsletter subscribers: Have you #UninstalledChatGPT and moved to Claude? 

~70% of respondents have switched or are considering switching to Claude.

The results were aligned with what we saw in the chat during the Profoundly kickoff last week. 

I have a follow-up question to this. Do you have Claude Max plan?

I operate at the intersection of AI, human genetics, and commercial strategy. At Genomics, I help pharmaceutical and biotech organizations bring large-scale human genetics into earlier, more confident R&D decision-making—through a combination of advanced analytics, AI-enabled tooling, and strategic consulting support.

My work centers on partnering with senior R&D, statistical genetics, and strategy leaders to operationalize genetics across the pipeline: accelerating target discovery, strengthening validation, improving translational confidence, and identifying patient populations more likely to respond to therapy.

Over the past 20 years, I’ve:

•	Built, led, and mentored high-performing sales teams in AI and SaaS.
•	Managed global accounts and developed go-to-market strategies that drive adoption.
•	Closed multi-million-dollar enterprise engagements and partnerships across R&D industries 
•	Negotiated strategic agreements that expand revenue and accelerate growth.

I’m also the author of Mastering AI Sales: Essential Knowledge for B2B Professionals, a practical playbook that equips founders, consultants, and sales leaders with frameworks to win complex AI sales.
👆 Now available – click the link in my profile header to get your copy.

Specialties

•	AI-Enabled Human Genetics / Genomics Commercial Strategy
•	Building & Scaling Sales Teams
•	Enterprise AI Sales & Stakeholder Management
•	Go-to-Market Strategy (Consulting + Product Hybrid Models)
•	Strategic Partnerships & Revenue Expansion
•	Cross-Functional Leadership & Commercial Alignment

Let’s connect if you’re scaling AI solutions, leading enterprise transformations, or looking to accelerate sales impact in complex markets.

David Walker

A striking article today about a potential brain drain in Massachusetts’ biomedical ecosystem.  Researchers are leaving academia, relocating to other states or countries, or exiting science altogether due to funding uncertainty and policy instability.

That’s concerning. Boston/Cambridge didn’t become the world’s leading biotech hub by accident. It took decades to build the flywheel between universities, hospitals, startups, and pharma.

But scientific ecosystems are fragile. When funding becomes unpredictable and talent leaves, the impact can last for years. In science, people are the engine of innovation.

The real question: How do we protect the ecosystems that produce the breakthroughs we all depend on?

For those of us working in research, genomics and drug discovery, this is something worth paying attention to. 

https://lnkd.in/eNP3MWGZ #DrugDiscovery #Boston #Innovation #Science

Commercial Leader at the Intersection of AI, Human Genetics & Pharma R&D | Strategic Partnerships | Enterprise AI Sales | Author | MBA MSc

Open article: ‘My scientific career is essentially over.’ A brain drain imperils Massachusetts’ biomedical future. by boston.com

‘My scientific career is essentially over.’ A brain drain imperils Massachusetts’ biomedical future.

There’s a misconception that AI will transform drug discovery primarily through better algorithms.

But several announcements over the past few days suggest something different. The real race is around the biological data scale.

Examples emerging this week:
• Machine-learning liquid biopsy approaches identifying disease signals from genome-wide circulating DNA fragmentation patterns
https://lnkd.in/gb9S79ic 
• Large biological foundation models trained on genomes from over 100,000 species, capable of predicting mutations and designing genetic sequences
https://lnkd.in/gE8ZzR-S 

The common thread is clear. AI performance in biology is increasingly determined by the depth, diversity, and integration of data.

This is why many pharmaceutical companies are investing heavily in:
• large-scale human genetics
• multimodal biological datasets
• clinical-genomic integration

In other words, the competitive advantage may not come from who builds the best model. It may come from who builds the best biological datasets.

#AI #DrugDiscovery #Genomics #Mystra #Biotech

One of the most interesting shifts in pharma R&D right now is the emergence of continuous learning loops between computational models and lab experiments.

Strategic investments are reinforcing this direction. A recent example is the $1B collaboration between NVIDIA and Eli Lilly and Company, aimed at building an AI factory for drug discovery, leveraging large-scale models trained on the language of biology and chemistry.

At the core of this approach is a tight feedback loop between the wet lab and the dry lab, where experimental results continuously update computational models.

Instead of the traditional discovery cycle:

Hypothesis → experiment → analysis → new hypothesis

AI enables something closer to:

Model prediction → experiment → real-time data → updated model → next experiment

This continuous loop allows research teams to iterate far more quickly.

Industry analyses suggest that embedding AI directly into experimental workflows could reduce discovery timelines by as much as 40% in some cases.

For pharma organizations, the implications are significant:

• accelerating target validation
• prioritizing experiments more effectively
• reducing failed experimental cycles

The companies that succeed may not simply use AI tools. They will build AI-native discovery systems in which computation and experimentation continuously inform one another.

Article:
https://lnkd.in/gzXqtj2Y

#AI #DrugDiscovery #Pharma #Biotech #PrecisionMedicine#AI #DrugDiscovery #Biotech #PharmaR&D

Open article: How AI is Transforming the Biopharmaceutical Value Chain from Discovery to Manufacturing | Pharmaceutical Technology by pharmtech.com

How AI is Transforming the Biopharmaceutical Value Chain from Discovery to Manufacturing | Pharmaceutical Technology

$500M is going into AI science, and not a single dollar is funding a new model.

A new nonprofit, Radial, just launched with that backing to build what most people overlook: the unglamorous infrastructure that AI-driven science actually runs on.

High-quality datasets. Standardized tools. Shared compute. Collaborative workflows.

Not headline-grabbing. Absolutely foundational.

It raises a question: If biology becomes a truly AI-driven discipline, will the most valuable asset in drug discovery be algorithms or the infrastructure that generates and integrates biological data at scale?

My bet: the companies that win the next decade of AI-driven drug discovery won't be the ones with the most sophisticated models. They'll be the ones who built the richest, most continuously learning biological data ecosystems.
The moat isn't the model. It's the data and infrastructure around it.

Where do you think the long-term advantage gets built, better models, better datasets, or both?

https://lnkd.in/ex5MzNwr
#AI #DrugDiscovery #Biotech #DataInfrastructure #AIStrategy

Open article: New nonprofit launches with at least $500 million to modernize scientific process for AI era by statnews.com

New nonprofit launches with at least $500 million to modernize scientific process for AI era

Human genetics is entering a new phase, one defined as much by data architecture as by biology.

Strong opening session this morning in the Integrated Precision Medicine track at PMWC on the opportunities in human genetics.

A few insights from the panel stood out:

Mark McCarthy (Genentech) discussed the opportunity to build population-scale models of health events, integrating genetics with layers such as proteomics and emerging immune signals like antibody profiles. A major challenge today is that many biobank datasets sit in isolated trusted research environments with limited interoperability. 

Mark Daly (Broad Institute) highlighted a fundamental challenge: most diseases involve hundreds to thousands of genetic variants, yet we still struggle to interpret what many of those variants actually do biologically. He also emphasized the need for lower-cost multi-omics assays and better interoperability across biobanks so datasets can be analyzed more easily across cohorts. 

Carlos Bustamante (University of Oklahoma) focused on the power of genetics to de-risk drug targets. But the scale required is enormous,  potentially ~10M sequenced genomes, with deep clinical data for ~1M individuals. Achieving this will require broader global infrastructure and participation across countries.

Nilah Ioannidis (UC Berkeley) emphasized that reference genomes alone are not enough. We need diverse datasets and richer functional layers, including transcriptomics and other omics data, to train better models of biology. Democratizing access to human genetics data will be key.

One theme that kept coming up:  The bottleneck in human genetics is increasingly not only data generation but making that data accessible, interoperable, and usable by scientists across R&D.

That shift toward scalable, accessible genetics infrastructure and the democratization of Human Genetics data is something we’re seeing more and more in conversations with Pharma teams as well.

Great start to PMWC.

#PMWC #HumanGenetics #PrecisionMedicine #Genomics #DrugDiscovery #AIinBiology

undefined undefined

Human genetics is no longer bottlenecked by sequencing.
It is bottlenecked by integration.

The scale of human genetics datasets is accelerating:
– UK Biobank: ~500,000 participants
– All of Us: >1 million participants
– Genomics England: 100,000+ genomes, with new population-scale efforts underway

On paper, this is what the field has been waiting for. 

But scale does not equal utility.

Much of this data still sits in siloed environments, under different access models, and across incompatible systems.

So, the paradox is clear:

We now have cohorts large enough to detect rare-variant signals, connect genetics to longitudinal outcomes, and train AI models on population-scale biology.

But we still struggle to use data across datasets, modalities, and workflows.
For pharma, that is not just a technical issue. It is a strategic one.

The winners will not be those with the most data.

They will be the ones who can make data sets interoperable and usable at speed and scale for organizational decision-making.

That is exactly the gap a new layer of infrastructure is starting to address.

That is also the problem we are focused on at Genomics. 

Is the next advantage in human genetics about owning more data, or making data usable across systems for faster decision-making at scale?

No alternative text description for this image

What an opening session for the AI track at the final day of the Precision Medicine World Conference (PMWC).

A fascinating fireside chat with two titans of the tech world:

Greg Brockman (Co-founder of OpenAI) and Vinod Khosla (Founder of Sun Microsystems and Khosla Ventures).

A few takeaways that really stood out:

1️⃣ The pace of AI progress is about to accelerate dramatically.
AI agents can work 24/7. AI is already improving laboratory work and automation. The only limitation is the compute is available. The rate of experimentation and iteration will become extraordinary. In the future will we need humans in the lab?

2️⃣ We may all become “CEOs of AI.”
In the future, individuals will orchestrate teams of AI agents, directing them to solve problems, run analyses, and generate insights. A new skill will be the ability to map AI models to business problems.

3️⃣ Science is becoming more democratized.
As AI tools improve, the barrier to doing sophisticated analysis will drop dramatically. Knowledge will be shared and applied far more seamlessly. We are still in the process of working out the feedback loop from AI to Humans.

4️⃣ Biology may become an AI-native domain.
DNA sequences are essentially structured data. With transformer-based models, AI can begin to learn biological systems in ways that may eventually predict biology better than humans can.

5️⃣ The bottleneck may shift.
If AI dramatically accelerates hypothesis generation, the constraints move elsewhere e.g. clinical trials, validation, and real-world experimentation.

6️⃣ The feedback loop between AI and humans is still evolving.

Current models have largely learned from academic data and software code. But as they interact more with the real world, their capabilities will expand quickly.

One striking thought from Greg: AI progress is like a rising tide, some mountains (hard problems) are already submerged, while others remain. But the water level is rising fast.

Vinod Khosla also touched on a provocative idea:  What if medicine eventually bypasses traditional clinical trials by moving toward deeply personalized, N=1 treatments?

And perhaps the biggest takeaway: When AI systems start meaningfully improving other AI systems, the rate of progress could increase dramatically.

We’re still early in this transition, but, as Greg quoted “it increasingly feels like a project of humanity to navigate the move into an AI-powered world”.

#AI #Genomics #PrecisionMedicine #PMWC #DrugDiscovery #MachineLearning

Another strong opening session on Day 2 of Precision Medicine World Conference 2026 in Santa Clara.

The panel in the Integrated Precision Medicine track explored a fundamental question:
How do we move from sequencing genomes to actually using them in patient care?

A few themes stood out:

🔹 Sequencing is still too expensive for routine clinical use
There was a lively discussion around who should ultimately pay for genetic diagnostics — insurers, healthcare providers, or pharmaceutical companies. An interesting question raised: will pharma eventually acquire diagnostic companies to control this capability?

🔹 Even countries leading in sequencing struggle with clinical adoption
Places like Finland and initiatives like FinnGen have generated enormous genomic datasets — yet much of this data is still not routinely used in clinical decision-making. The challenge isn’t just data generation; it’s integration into healthcare systems.

🔹 AI is coming rapidly into clinical medicine — but people are the constraint
Jeffrey Balser (Vanderbilt University Medical Center) emphasized that while AI adoption in healthcare will accelerate, organizational change and people will be the biggest challenge.

🔹 Integrating research and clinical data and using the data is a strategic priority
Howard Jacob (AbbVie) described efforts to bring R&D and clinical data into a unified data warehouse, alongside investment in companion diagnostics (CDx).

🔹 Diagnostics remain undervalued relative to therapeutics
Euan Ashley (Stanford University) pointed out that the market capitalization of pharmaceutical companies is roughly 3× that of diagnostics companies, despite diagnostics being foundational to precision medicine.

A few items from the “wish list” for the future of precision medicine:

• Clear reimbursement pathways for diagnostics
• AI-driven clinical diagnostic systems that integrate directly with health records that make clinicians life’s easier. 
• A deeper biological understanding of what truly defines disease, and using that knowledge in the clinic

The scientific capability is advancing rapidly.
The real challenge now is turning genomic data into routine clinical infrastructure.

AI is not just changing drug discovery.
It’s changing how scientific organizations operate.

After several recent conversations with pharma R&D teams, one thing is becoming clear.

The biggest impact of AI may not be better siloed models. 

It may be the emergence of AI-native research workflows.

Scientists are increasingly working alongside:

• AI copilots that generate hypotheses from literature and datasets
• AI agents running analyses across large biological data platforms
• automated laboratories executing experiments continuously

In that environment, the role of scientists starts to shift.

Less time running individual analyses.
More time orchestrating systems that generate and test ideas at scale.

That has major implications for how R&D organizations are structured.

Instead of traditional lab teams organized around projects, we may see research groups evolve into AI-orchestrated discovery systems combining:

• biological data infrastructure
• foundation models
• automated experimentation platforms

In other words, the transformation may not just be technological.

It may be organizational.

The long-term shift could be subtle but profound:

Drug discovery organizations may start to look less like traditional labs
and more like AI-driven engineering systems.

Which raises an interesting question:

Will pharma R&D start to resemble software development teams in the next decade?

#AI #DrugDiscovery #Biotech #Genomics #Pharma

Just over three months in at Genomics and what a journey! I feel fortunate to work with a team committed to advancing the use of human genetics in R&D and delivering real impact for companies pursuing novel therapeutics.

A few early observations:

1. There is an enormous appetite across pharma to use human genetics more effectively, from target identification to patient stratification and POS-informed decisions. Yet many organizations are constrained not by scientific expertise, but by capacity, infrastructure burden, and legacy statistical genetics tools that don’t scale with therapeutic area demand.

2. Statistical geneticists are frustrated, spending too much time on manual analysis, data curation, and QA. Seeing workflows that once took days executed in under a minute during Mystra demos has been eye-opening.

3. An important shift is the focus on the democratization of data. With agentic AI integrated into Mystra, structured genetic insight can extend beyond the statistical genetics function to translational and portfolio stakeholders, without compromising scientific rigor.

Human Genetics teams are under sustained pressure. Demand continues to rise, but the inefficiencies and bottlenecks don’t have to persist.

I’ll be at the Precision Medicine World Conference in Santa Clara next week. If this resonates, I’d welcome the opportunity to connect and continue the conversation.

pmwcintl.com

PMWC 2027 | Precision Medicine World Conference, Jan. 27-29

A major milestone for Genomics and an important one for statistical genetics teams.

Human Genetics teams are under sustained pressure to support more programs and increasingly complex analyses, without proportional growth in headcount. The constraint isn’t scientific expertise it’s access to harmonized data, capacity and coordination.

Agentic AI within Mystra directly addresses those constraints.
Its integration represents a material shift in how statistical genetics workflows are executed. Rather than manually orchestrating pipelines, harmonizing datasets, and iterating through hypothesis tests, statistical geneticists can delegate complex, multi-step analyses to AI agents that reason across data, execute workflows, and return structured, decision-ready outputs.

The result is the ability to absorb portfolio-level demand while maintaining analytical depth, reproducibility, and decision quality.

For Human Genetics groups supporting multiple therapeutic areas, the implications are significant:

• Faster target prioritization
• Reduced infrastructure and orchestration burden
• Fewer analytical bottlenecks
• More consistent, portfolio-ready evidence packages

Excited to see this launched from the new London office.

📢Genomics is delighted to announce that Dr Zubir Ahmed MP, the UK Government Minister for Health Innovation & Safety, and Steve Bates OBE FMedSci, Executive Chair, Office of Life Sciences, today open our flagship London office. The opening marks our formal membership within the Knowledge Quarter community in Kings Cross.

At the office opening, Genomics will unveil new agentic AI capabilities integrated into the Mystra platform. Mystra offers biopharma partners proven analytical power to turn months of R&D into minutes, dramatically increasing the availability of quality genetic insights and, therefore, the probability of honing in on successful drug candidates.

📰Read the full press release here: https://lnkd.in/ee7u7MDc

#Genomics #Innovation #UKInnovation #Mystra #LifeSciences #PrecisionMedicine #News #CompanyNews #IndustryNews

Open article: Minister for Health Innovation Dr Zubir Ahmed MP and Executive Chair for the Office of Life Sciences Steve Bates OBE open Genomics’ flagship office in London’s Knowledge Quarter, as it unveils Agentic AI capabilities within Mystra drug discovery Platform | Genomics by genomics.com

Minister for Health Innovation Dr Zubir Ahmed MP and Executive Chair for the Office of Life Sciences Steve Bates OBE open Genomics’ flagship office in London’s Knowledge Quarter, as it unveils Agentic AI capabilities within Mystra drug discovery Platform | Genomics

Roche’s NVIDIA announcement is notable. But the real signal is bigger than one company buying GPUs.

Roche has launched a large-scale NVIDIA-powered AI factory, adding 2,176 Blackwell GPUs and taking its combined on-prem and cloud footprint to more than 3,500 GPUs. Roche says this is intended to support AI across the value chain, from discovery and development to manufacturing and diagnostics. 

To me, this marks an important shift in how large pharma is approaching AI.
For the past few years, much of the conversation in biotech has focused on models and algorithms.

But the emerging competitive question may be different:

Who is building the compute, data, and workflow infrastructure required to make AI usable at scale in biology? That is a harder problem than model selection.

In drug discovery, advantage will not come from GPUs alone.
It will come from combining:
• large-scale compute
• proprietary biological and clinical data
• lab-in-the-loop workflows
• the ability to move from model output to experimental validation quickly

That is why announcements like this matter.

They suggest some pharma companies are no longer treating AI as a set of tools layered onto R&D. They are starting to treat it as core scientific infrastructure.

And that may end up being the more important shift.

The companies that build AI-native discovery infrastructure, not just AI experiments, may be the ones that compound advantage over the next 5–10 years.

https://lnkd.in/gg8TBpAT

#AI #DrugDiscovery #Biotech #ComputationalBiology #Pharma

Open article: Roche inks Nvidia deal to bolster AI factory, speed up drug and diagnostic development by fiercebiotech.com

AI, artificial intelligence, factory, cloud

Roche inks Nvidia deal to bolster AI factory, speed up drug and diagnostic development

The biggest AI workforce shift in pharma may not be job loss. It may be job redesign.

Pharma may be spared the immediate AI-driven layoffs seen in parts of tech.
That does not mean pharma is spared from workforce disruption. 

BioPharma Dive made the case that biopharma is less exposed to near-term AI replacement than sectors where work is easier to standardize. 

That feels directionally right. Drug discovery and development still depend on scientific judgment, messy biological data, regulated processes, and human interpretation. 

But “less exposed” is not the same as unchanged.

Across the broader economy, companies are already linking layoffs to AI efficiency pushes, and Reuters reported AI was tied to 7% of announced U.S. layoffs in January. 

Pharma’s near-term shift is likely different:

-More hybrid scientific/computational roles.
-More internal AI and analytics capability.
-More value is placed on people who can translate models into real R&D decisions. 

The real question for leaders is not whether AI will reduce headcount. It is whether they are redesigning roles early enough for an AI-enabled R&D model.

Open article: AI is slashing jobs across industries. Will pharma be next? by biopharmadive.com

AI is slashing jobs across industries. Will pharma be next?

One of the clearest signals in biotech right now is the rise of AI partnerships across the industry.

A recent collaboration between Insilico Medicine and Liquid AI aims to deploy lightweight AI models to accelerate different stages of the drug discovery process.

This reflects a broader shift. Instead of building everything internally, companies are increasingly combining strengths across:

• AI model development
• biological datasets
• experimental validation

Drug discovery is becoming less of a single-company capability and more of an interdisciplinary platform problem.

No single organization typically has:

• the best AI models
• the largest biological datasets
• the deepest disease expertise
• the most advanced experimental infrastructure

Strategic partnerships allow these capabilities to come together. The future of drug discovery may look less like isolated research programs and more like collaborative technology ecosystems.

#AI #DrugDiscovery #Genomics #Mystra #Biotech #Pharma #DigitalBiology

Open article: Partnership brings new AI model promising faster drug discovery by longevity.technology

Partnership brings new AI model promising faster drug discovery

taplio

The Battle

Klemen Hrovat

David Walker

Followers

Posts

Engagements

Likes

Performance ScoreKlemen Hrovat's LinkedIn Performance

Performance ScoreDavid Walker's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Klemen Hrovat

David Walker

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Klemen Hrovat's LinkedIn Performance

Performance ScoreDavid Walker's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreKlemen Hrovat's LinkedIn Performance