Prashant Jalan launches Guickly to find where enterprise AI budgets went

Why it matters

Guickly is betting that AI cost control will become a software category of its own as subscriptions, APIs and autonomous agents turn enterprise AI into a metered expense that finance cannot track with SaaS-era tools.

Prashant Jalan launched Guickly publicly on September 1st with software designed to show enterprises which AI tools, agents and models their employees use, what those systems cost and whether the spending produces a measurable return. Guickly also announced a $4.2M seed round led by Engineering Capital, with participation from Converge VC, Neon Fund and angel investors.

The launch turns Jalan's experience measuring computing performance inside Google into a broader bet on enterprise finance. Jalan spent eight years at Google and most recently led applied AI work in Google Maps, according to Guickly. Guickly credits him with helping build Maps' speed-limit feature and developing a profiler for optimizing Google's tensor processing units.

"When I was leading an Applied AI team at Google and making products for billions of people, I traced where every cycle and every byte was going," Jalan said in the September 1st announcement. He said that experience led him to a problem beyond model performance: large organizations could not readily account for the AI operating inside their own walls.

The financing predates the launch

September 1st marks Guickly's public debut, rather than the beginning of its financing. An amended Form D filed with the SEC lists November 14th, 2025 as the offering's first sale date. The filing, signed by Jalan on May 18th, 2026, records $4,234,993 sold from a planned $5M equity offering to seven investors.

That amount closely matches the $4.2M announced by Guickly. The SEC record also shows Guickly was incorporated in Delaware in 2025 and lists San Jose, California, as its principal place of business. Guickly did not announce a valuation.

The timeline suggests Jalan raised the capital while building in stealth, then waited to introduce Guickly until the product and its sales pitch were ready for enterprise buyers. Guickly says it developed the system with input from CIOs and other senior executives and is working with early design partners and customers. Guickly has not published customer names, measured savings or deployment results.

Engineering Capital's investment fits the firm's stated practice of leading seed rounds built around a founder's technical insight. Ashmeet Sidana, Engineering Capital's chief engineer, said he had followed Jalan's career for years and backed his ability to turn technical experience into a commercial product.

Turning AI usage into a finance problem

Guickly's core argument is that AI has broken the purchasing model enterprise software departments spent the past decade learning to manage. Conventional SaaS costs are often tied to a known number of seats. AI costs can accumulate through subscriptions, model calls, internal agents, cloud accounts, browser tools and personal accounts used for work.

That produces two related problems. Finance sees bills without a clean explanation of which departments or projects generated them. Security and IT may have an incomplete inventory because employees can adopt AI products without going through centralized procurement.

Guickly calls the resulting mess "token sprawl." Its product is meant to create a live inventory of AI tools, models, agents and vendors, then attribute subscriptions and API expenses to individual users, departments and applications. Guickly says administrators can set budgets and policies, identify unused licenses, flag overlapping products and route workloads toward less expensive models.

The monitoring extends beyond tools purchased by procurement. Guickly says it detects web applications, desktop clients, API calls, browser extensions and plugins used inside development environments. That breadth matters because a license-management dashboard cannot account for an engineer charging model calls to a cloud project or an employee using a personal AI account with corporate data.

Jalan is also selling privacy as part of the architecture. Guickly says prompts, responses, source code and other sensitive content remain on premises, while Guickly works with metadata. Guickly's current site says the product requires no SDK, code changes, instrumentation or orchestration, and that customers can go live in hours. Those deployment and privacy claims have yet to be demonstrated through public customer case studies.

The approach reflects Jalan's preference for eliminating the work required to assemble an answer. "We believe time is sacred, and speed is respect," he said in the launch announcement. Guickly's founder manifesto develops that idea around the finite number of weeks in a person's life and the time executives lose compiling reports or scheduling meetings to find a number.

AI cost control is already getting crowded

Guickly enters a category that is forming quickly around AI observability, financial operations and governance. CloudZero launched an AI financial control plane on May 28th, extending its cloud-cost platform to connect model activity with products, customers and business outcomes. Fiddler AI raised a $30M Series C earlier in 2026 for a broader control plane covering AI monitoring, evaluation, security and governance.

Guickly is positioning itself across several of those functions at once: discovering sanctioned and unsanctioned AI, allocating costs, measuring adoption, enforcing budgets and estimating returns. The on-premises processing model could help Jalan sell into finance, pharmaceutical and other regulated customers, where sending prompts or source code to another vendor creates a separate approval problem.

The breadth also raises the execution bar. Discovering an AI service is easier than proving the revenue, savings or productivity it produced. McKinsey's 2026 global AI survey found that 37% of respondents attributed at least some EBIT impact to AI use, while about 6% of all respondents qualified as high performers that attributed at least 5% of EBIT to AI and described its impact as significant.

Guickly will need reliable data from finance, procurement, identity systems, developer infrastructure and business workflows to close that attribution gap. It must do so without collecting the sensitive content that could provide useful context. Jalan's profiler experience gives Guickly a credible technical starting point. The commercial test is whether executives will trust one new dashboard to reconcile an AI bill assembled across dozens of old ones.