The venture firms betting on AI are now using it to rethink their own ops and due diligence.
The AI gold rush is at its peak in 2025. Every day, a new AI startup raises funds with the promise to change the world, or a prominent AI lab raises a massive round at an eye-popping valuation, with the backing of marquee venture funds.
In the first half of this year alone, more than half of all global venture capital flowed into AI companies, up from just 25-30% a year earlier. This investor appetite only appears to be growing.
However, with this trend, a new question is emerging: how are these angels and VCs, who are betting so heavily on AI, actually using it themselves?
“There are three different impacts [of AI]. The quality of due diligence expected is significantly higher, the speed of that diligence is much shorter, and the structuring of diligence across a firm needs to be pristine,” Tomasz Tunguz, founder of Theory Ventures, which recently hired a Head of AI to implement internal AI workflow initiatives, said in an interview with Future Nexus.
Tunguz is one of the many VCs who are using AI today to automate portions of their investment workflows, which have largely remained manual over the years.
The new VC playbook
For any VC, the key to successful business is a successful exit, which means they have to find the right company from a pool of thousands of potential candidates — one that is legitimate and could grow well and fast enough to make them money.
To undertake this herculean task, the firm must source potential candidates using networks, conferences, and databases like PitchBook or Crunchbase, then cut through the noise via a deep understanding of parameters like competitors, user feedback, growth metrics, and team background information.
“A single analyst might review 300-500 companies per quarter and still miss critical outliers. Due diligence requires weeks of manual market sizing, founder interviews, and reference calls, and memos crafted from scratch,” angel investor Sandeep Kondury told Future Nexus.
The effort required through this entire process could often delay the investment, sometimes even leading to missed opportunities.
But, not anymore.
With AI tools in the pipeline, every step is being re-engineered, giving VCs the ability to automate a lot of their work.
“Scouting now uses intelligent filters that surface the top 2–3% of deals relevant to a fund’s thesis, reducing noise by up to 70%,” Kondury said while explaining the shift in startup research and due diligence.
He said that AI can now perform automated market mapping, founder language analysis, and sentiment mining to cut the research time by as much as 60%, condensing the whole process from several weeks to just days. And it further helps with operational processes and portfolio allocation optimization.
“At the operational level, we use ChatGPT in agent mode for routine but time-consuming tasks — calendar prep, call summaries, even first-pass market maps. For investment work, we lean on ChatGPT for deal underwriting and memo drafting; what once took 12-15 analyst hours can now be reduced to 2-3. For execution and pipeline management, Comet has become a favorite. We’ve seen about a 40% increase in response rates on cold outreach when AI curates messaging around sector-specific language,” Kondury noted.
For the most VC-intensive workflows like portfolio optimization, he uses Decile Hub, a dedicated platform for VC ops.
“It allows us to create internal decision docs, model fund outcomes, and benchmark portfolio companies in a fraction of the time. What used to take an investment team days of manual spreadsheet work is now condensed into hours, enabling us to review 2-3x more opportunities per quarter without expanding headcount,” he added.
Human edge still matters
While both Tunguz and Kondury emphasized that AI is enabling them to automate as much of the analytical work as possible, they didn’t refrain from highlighting some glaring caveats.
First, Tunguz pointed out that even with AI tools at their disposal, they’re not doing away with “human judgment,” with the final call still staying in the hands of a human, be it during the research or portfolio allocation phase.
“[The goal] isn’t to automate judgment itself, but to automate many of the diligence functions involved in competitive analysis,” Tunguz told Future Nexus.
Secondly, Kondury noted that no matter what tool they use, the context remains the key to success with AI.
“We’ve focused heavily on embedding our firm’s thesis and unique pattern-recognition framework into the tools we use. When tuned correctly, they sharpen decision quality while providing the benefits of speed,” he said.
If not tuned correctly, the AI would be nothing more than a general-purpose chatbot and may even highlight inaccurate aspects of a deal. This is also where human judgment plays an important role, as one has to decide what to train the AI to look for, what to ignore, and how to interpret the patterns it surfaces. In other words, investors move up the stack, from data processors to thesis architects.
Both investors also highlighted the value of balancing between open and closed source models. Essentially, teams should focus on closed models (with maximum privacy settings) when speed and performance are crucial.
However, when there’s room to tone down performance a bit or the workflow is way too sensitive, open-source with self-hosted environments should be the way to go for full data control and governance.
“Our policy is simple: no confidential deal data in black-box systems. The future is hybrid: firms that secure and fine-tune their own models will have both founder trust and a durable competitive edge,” Kondury said.
A productivity curve like never before
When used wisely and with full privacy control, AI’s efficiency gains are hard to ignore. Kondury estimates that it is shifting the industry from being reactive to proactive, enabling partners to effectively oversee 2-3x more pipeline — all this, while also increasing conviction and providing them live portfolio data using simple text prompts.
“Soon, we plan to build an AI layer to automate our inbound application process, screening startups against our thesis to drastically reduce false positives. In the next two years, we expect this to evolve into a fully adaptive system: AI agents that not only filter but also engage founders with tailored feedback, dynamically track market shifts, and flag new opportunities in real time. The aim is simple: shift analyst hours from manual screening to high-leverage judgment, where human and machine intelligence compound instead of compete,” he said.
Tunguz similarly shared that as AI increasingly automates research, the role of an investor will lean more heavily on relationship building and judgment.
Simply put, the next decade of venture capital won’t be decided by who has the deepest pockets, but by who masters the balance of human intuition and machine precision.
AI will help VCs triple their pipeline, but speed should always be accompanied by signal. The firms that will use AI to sharpen — not replace — human judgment will come out as the real winners, defining which founders rise, which products endure, and which markets truly matter.
