Tarun Mathur
@tarunmathur.bsky.social
#GrowthAdvisory #CorporateStrategy #ScaleupStrategy #FoundersAdvice #Fintech #AI #Web3 #EnterpriseTech #Cybersecurity | #KelloggAlum
In 2026, expect AI asset management to become as standard as CMDBs for infrastructure.
That means building inventories of models, agents, prompts, datasets, and integrations across the enterprise.
#TMinsights
That means building inventories of models, agents, prompts, datasets, and integrations across the enterprise.
#TMinsights
January 10, 2026 at 10:07 AM
In 2026, expect AI asset management to become as standard as CMDBs for infrastructure.
That means building inventories of models, agents, prompts, datasets, and integrations across the enterprise.
#TMinsights
That means building inventories of models, agents, prompts, datasets, and integrations across the enterprise.
#TMinsights
Many organizations still treat AI as an incremental add-on. Ideally, they should use AI to challenge the shape of org charts, processes, and even business models.
That mindset shift is the real strategic unlock.
#TMinsights
That mindset shift is the real strategic unlock.
#TMinsights
January 10, 2026 at 9:21 AM
Many organizations still treat AI as an incremental add-on. Ideally, they should use AI to challenge the shape of org charts, processes, and even business models.
That mindset shift is the real strategic unlock.
#TMinsights
That mindset shift is the real strategic unlock.
#TMinsights
Generative AI was act one. Act two is agentic AI—autonomous systems that don’t just generate; they strategize, plan, and execute multi-step tasks without constant human intervention.
Your competitive advantage won’t come from chatbots. It’ll come from orchestrated agent networks.
#TMinsights
Your competitive advantage won’t come from chatbots. It’ll come from orchestrated agent networks.
#TMinsights
January 4, 2026 at 6:23 AM
Generative AI was act one. Act two is agentic AI—autonomous systems that don’t just generate; they strategize, plan, and execute multi-step tasks without constant human intervention.
Your competitive advantage won’t come from chatbots. It’ll come from orchestrated agent networks.
#TMinsights
Your competitive advantage won’t come from chatbots. It’ll come from orchestrated agent networks.
#TMinsights
Stop the "pray and spray" pilot spending. It's time to build measurement rigor into every AI initiative.
#TMinsights
#TMinsights
January 4, 2026 at 6:20 AM
Stop the "pray and spray" pilot spending. It's time to build measurement rigor into every AI initiative.
#TMinsights
#TMinsights
The next generation of identity is cryptographically provable, consent-driven, and privacy-shielded.
#TMinsights
#TMinsights
January 2, 2026 at 1:05 PM
The next generation of identity is cryptographically provable, consent-driven, and privacy-shielded.
#TMinsights
#TMinsights
January 2, 2026 at 9:26 AM
The cost of a data, privacy, or cyber violation is rarely just the fine; it’s the opportunity cost of delayed deals, lost partnerships, and the intangible erosion of brand equity.
Invest in a holistic AI governance stack—observability, consent management, ethics boards.
#TMinsights
Invest in a holistic AI governance stack—observability, consent management, ethics boards.
#TMinsights
January 1, 2026 at 4:38 PM
The cost of a data, privacy, or cyber violation is rarely just the fine; it’s the opportunity cost of delayed deals, lost partnerships, and the intangible erosion of brand equity.
Invest in a holistic AI governance stack—observability, consent management, ethics boards.
#TMinsights
Invest in a holistic AI governance stack—observability, consent management, ethics boards.
#TMinsights
The fusion of AI with BFSI promises unprecedented efficiency, but it also disrupts the domains of identity, privacy, consent, and cyber ethics.
Your focus and strategies will determine whether these pillars are upheld through robust governance or violated in the rush for innovation.
#TMinsights
Your focus and strategies will determine whether these pillars are upheld through robust governance or violated in the rush for innovation.
#TMinsights
January 1, 2026 at 3:25 PM
The fusion of AI with BFSI promises unprecedented efficiency, but it also disrupts the domains of identity, privacy, consent, and cyber ethics.
Your focus and strategies will determine whether these pillars are upheld through robust governance or violated in the rush for innovation.
#TMinsights
Your focus and strategies will determine whether these pillars are upheld through robust governance or violated in the rush for innovation.
#TMinsights
The new leadership competency isn't just deploying AI—it's explaining it. This is where technical leadership meets storytelling.
Future leaders must interrogate models, understand confidence intervals, detect model drift, and decide when to override AI with human judgment.
#TMinsights
Future leaders must interrogate models, understand confidence intervals, detect model drift, and decide when to override AI with human judgment.
#TMinsights
December 30, 2025 at 9:34 AM
The new leadership competency isn't just deploying AI—it's explaining it. This is where technical leadership meets storytelling.
Future leaders must interrogate models, understand confidence intervals, detect model drift, and decide when to override AI with human judgment.
#TMinsights
Future leaders must interrogate models, understand confidence intervals, detect model drift, and decide when to override AI with human judgment.
#TMinsights
December 30, 2025 at 9:31 AM
The 2025 plan: hire a handful of AI experts and hope they transform the company.
The 2026 plan: turn every role into an AI-augmented role.
This shift—from specialist-centric to workforce-centric AI—is how enterprises can escape the talent bottleneck.
#TMinsights
The 2026 plan: turn every role into an AI-augmented role.
This shift—from specialist-centric to workforce-centric AI—is how enterprises can escape the talent bottleneck.
#TMinsights
December 30, 2025 at 9:18 AM
The 2025 plan: hire a handful of AI experts and hope they transform the company.
The 2026 plan: turn every role into an AI-augmented role.
This shift—from specialist-centric to workforce-centric AI—is how enterprises can escape the talent bottleneck.
#TMinsights
The 2026 plan: turn every role into an AI-augmented role.
This shift—from specialist-centric to workforce-centric AI—is how enterprises can escape the talent bottleneck.
#TMinsights
Most enterprises say data quality is their #1 blocker to scaling AI. These are not data‑poor companies—they’re data‑rich and context‑poor.
You can’t out‑model broken data. You have to redesign pipelines, ownership, and governance or the models will simply expose your mess faster.
#TMinsights
You can’t out‑model broken data. You have to redesign pipelines, ownership, and governance or the models will simply expose your mess faster.
#TMinsights
December 30, 2025 at 9:03 AM
Most enterprises say data quality is their #1 blocker to scaling AI. These are not data‑poor companies—they’re data‑rich and context‑poor.
You can’t out‑model broken data. You have to redesign pipelines, ownership, and governance or the models will simply expose your mess faster.
#TMinsights
You can’t out‑model broken data. You have to redesign pipelines, ownership, and governance or the models will simply expose your mess faster.
#TMinsights
Consumer-style prompts don’t translate to mission-critical, multi-step processes.
Architect for that gap or pay for it in production.
#TMinsights
Architect for that gap or pay for it in production.
#TMinsights
December 30, 2025 at 8:21 AM
Consumer-style prompts don’t translate to mission-critical, multi-step processes.
Architect for that gap or pay for it in production.
#TMinsights
Architect for that gap or pay for it in production.
#TMinsights
December 27, 2025 at 3:27 AM
The most concerning metric from the year-end reports I'm reading is: “Atrophy of critical thinking skills due to GenAI.”
We need to retain focus on “First Principles Thinking.”
If you can’t think without Copilot, you’re a liability in a crisis.
#TMinsights
We need to retain focus on “First Principles Thinking.”
If you can’t think without Copilot, you’re a liability in a crisis.
#TMinsights
December 26, 2025 at 4:28 AM
The most concerning metric from the year-end reports I'm reading is: “Atrophy of critical thinking skills due to GenAI.”
We need to retain focus on “First Principles Thinking.”
If you can’t think without Copilot, you’re a liability in a crisis.
#TMinsights
We need to retain focus on “First Principles Thinking.”
If you can’t think without Copilot, you’re a liability in a crisis.
#TMinsights
The “Agent Marketplace” concept is evolving faster than SaaS did.
The competitive advantage shifts from building AI to assembling agent teams.
The competitive advantage shifts from building AI to assembling agent teams.
December 26, 2025 at 4:22 AM
The “Agent Marketplace” concept is evolving faster than SaaS did.
The competitive advantage shifts from building AI to assembling agent teams.
The competitive advantage shifts from building AI to assembling agent teams.
The shift from “One Giant Model” to “Router Networks” (routing queries to the cheapest/fastest model) will be the way to save on compute costs.
2026 could be the year of the “Model Router.”
Optimization is the new innovation.
2026 could be the year of the “Model Router.”
Optimization is the new innovation.
December 26, 2025 at 3:59 AM
The shift from “One Giant Model” to “Router Networks” (routing queries to the cheapest/fastest model) will be the way to save on compute costs.
2026 could be the year of the “Model Router.”
Optimization is the new innovation.
2026 could be the year of the “Model Router.”
Optimization is the new innovation.
Deep Domain Expertise + Commodity AI >> Shallow Expertise + Commodity AI.
December 26, 2025 at 3:20 AM
Deep Domain Expertise + Commodity AI >> Shallow Expertise + Commodity AI.
Enterprises don’t want to pay a large LLM for every internal query. They want a “good enough” model that runs cheap and fast.
This will result in us seeing more model distillation: train on a massive frontier model, then distill the knowledge into a tiny model.
2026 could see "corporate SLM.”
This will result in us seeing more model distillation: train on a massive frontier model, then distill the knowledge into a tiny model.
2026 could see "corporate SLM.”
December 25, 2025 at 5:13 PM
Enterprises don’t want to pay a large LLM for every internal query. They want a “good enough” model that runs cheap and fast.
This will result in us seeing more model distillation: train on a massive frontier model, then distill the knowledge into a tiny model.
2026 could see "corporate SLM.”
This will result in us seeing more model distillation: train on a massive frontier model, then distill the knowledge into a tiny model.
2026 could see "corporate SLM.”
Prompt engineering is essential. But equally important is “agent management.”
We should be able to evaluate the AI’s reasoning, and intervene when it deviates.
However, in a world where so many professionals lack effective management skills, upskilling in logic and delegation is critical.
We should be able to evaluate the AI’s reasoning, and intervene when it deviates.
However, in a world where so many professionals lack effective management skills, upskilling in logic and delegation is critical.
December 25, 2025 at 2:40 PM
Prompt engineering is essential. But equally important is “agent management.”
We should be able to evaluate the AI’s reasoning, and intervene when it deviates.
However, in a world where so many professionals lack effective management skills, upskilling in logic and delegation is critical.
We should be able to evaluate the AI’s reasoning, and intervene when it deviates.
However, in a world where so many professionals lack effective management skills, upskilling in logic and delegation is critical.
We spent the past decades securing “users” (identity management).
Now we have to secure “agents.”
When an AI agent has permission to read your email, query the database, and send a Slack message, it is an insider threat by design.
CISOs should start talking about “non-human identity management.”
Now we have to secure “agents.”
When an AI agent has permission to read your email, query the database, and send a Slack message, it is an insider threat by design.
CISOs should start talking about “non-human identity management.”
December 25, 2025 at 2:17 PM
We spent the past decades securing “users” (identity management).
Now we have to secure “agents.”
When an AI agent has permission to read your email, query the database, and send a Slack message, it is an insider threat by design.
CISOs should start talking about “non-human identity management.”
Now we have to secure “agents.”
When an AI agent has permission to read your email, query the database, and send a Slack message, it is an insider threat by design.
CISOs should start talking about “non-human identity management.”
As AI coding tools become widespread, models are effectively replacing mid-level software engineers.
This can lead to “software inflation”: if code is cheap, we will build more software.
Every department might be building its own mini-apps and agents.
The CIO becomes “governor of chaos.”
This can lead to “software inflation”: if code is cheap, we will build more software.
Every department might be building its own mini-apps and agents.
The CIO becomes “governor of chaos.”
December 25, 2025 at 2:09 PM
As AI coding tools become widespread, models are effectively replacing mid-level software engineers.
This can lead to “software inflation”: if code is cheap, we will build more software.
Every department might be building its own mini-apps and agents.
The CIO becomes “governor of chaos.”
This can lead to “software inflation”: if code is cheap, we will build more software.
Every department might be building its own mini-apps and agents.
The CIO becomes “governor of chaos.”
Your employees aren’t waiting for IT to approve a tool. They are using their own $20/month subscriptions to automate their jobs.
This “Shadow AI” creates a massive governance hole. But it also creates a massive signal.
Don’t ban it. Map it.
That is your roadmap for enterprise adoption.
This “Shadow AI” creates a massive governance hole. But it also creates a massive signal.
Don’t ban it. Map it.
That is your roadmap for enterprise adoption.
December 25, 2025 at 1:48 PM
Your employees aren’t waiting for IT to approve a tool. They are using their own $20/month subscriptions to automate their jobs.
This “Shadow AI” creates a massive governance hole. But it also creates a massive signal.
Don’t ban it. Map it.
That is your roadmap for enterprise adoption.
This “Shadow AI” creates a massive governance hole. But it also creates a massive signal.
Don’t ban it. Map it.
That is your roadmap for enterprise adoption.
The decoupling of revenue from “seats” is the biggest disruption to the B2B business model in the past two decades.
If you're a SaaS leader, you need to figure out how to cannibalize your own seat-based revenue before an agentic startup does it for you.
If you're a SaaS leader, you need to figure out how to cannibalize your own seat-based revenue before an agentic startup does it for you.
December 25, 2025 at 1:37 PM
The decoupling of revenue from “seats” is the biggest disruption to the B2B business model in the past two decades.
If you're a SaaS leader, you need to figure out how to cannibalize your own seat-based revenue before an agentic startup does it for you.
If you're a SaaS leader, you need to figure out how to cannibalize your own seat-based revenue before an agentic startup does it for you.
The moats of 2026 aren’t in the model weights; they’re in workflow integration.
If you’re a founder, don’t build a “better LLM.” Build a “better lawyer,” “better doctor assistant,” or “better underwriting engine.”
If you’re a founder, don’t build a “better LLM.” Build a “better lawyer,” “better doctor assistant,” or “better underwriting engine.”
December 25, 2025 at 1:20 PM
The moats of 2026 aren’t in the model weights; they’re in workflow integration.
If you’re a founder, don’t build a “better LLM.” Build a “better lawyer,” “better doctor assistant,” or “better underwriting engine.”
If you’re a founder, don’t build a “better LLM.” Build a “better lawyer,” “better doctor assistant,” or “better underwriting engine.”