Olakai

• Enterprises could fund AI cost control from engineering or finance budgets and workforce AI governance from security budgets, letting specialists such as DX on one side and Harmonic Security or WitnessAI on the other take each half of Olakai's offer.
• Microsoft, the model providers and the agent platforms could ship native usage and return dashboards good enough for their own tools, shrinking the cross-vendor gap Olakai charges for.
• The four testimonials on Olakai's homepage are all anonymous. If no named reference customer appears by mid-2027, a buyer still has only the vendor's word for traction.
• Olakai's own pages count the tools it covers as 850 or more in one place, 800 or more in another, and 630 or more in a third. A pilot that finds discovery coverage materially below the headline number would undercut the governance half of the pitch.
• AI Fund backs both Olakai and one of its two named partners, so one of the two named references sits inside the investor's network.

Olakai sells to the executives who have to justify enterprise AI spending. The homepage states that most leaders cannot prove what their AI spend is returning and that the subsidized era is over, with coding-agent costs up five to ten times, and use-case pages address the head of AI, the CFO, the CISO and the VP of engineering separately. Each of those buyers meets a different alternative, which is why the pitch splits four ways before the product does.

The pain argument is sharpest on engineering spend. Olakai's coding product page describes forecasting month-end spend from a run-rate and firing an alert before the invoice arrives, and the homepage carries a quote from an unnamed engineering leader whose token bill grew fivefold in a quarter after Claude Code moved to usage-based pricing. Every one of those voices is a role without a company behind it, so a buyer cannot trace the pain Olakai describes to an organization that reported it.

Outside voices confirm the question rather than the size of it. HP Tech Ventures' February 2026 feature treats the unanswered question of whether AI investment is working as common across companies, and Olakai's own manifesto reaches for a PwC survey of more than 4,000 chief executives in which 56% report zero financial benefit. That figure is the survey's, carried on Olakai's page, and no independent research in the reviewed sources measures the pain for this product class. [s1, s3, s8, s22, s12]

Olakai now sells two products on one data layer. The platform page states that Olakai includes Olakai Agentic, for the return on AI coding tools and the governance of autonomous agent workflows, and Olakai Assistive, for adoption and return across copilots, chat assistants and AI features inside other software. Underneath them sit three measurement modules and Kai, a conversational layer that answers questions across all of them.

The measurement methods are described rather than asserted. The coding product classifies every pull request as AI-assisted or not by reading commit co-author trailers, bot authorship and body markers, then compares cycle time against the team's own history. A separate measure Olakai calls AI Equivalent Engineers converts the change in code shipped per week after adoption into dollars against a loaded engineer cost. Olakai's documentation states that assistive value comes from estimating the minutes of human work each interaction likely saved, which is an estimate the buyer has to accept rather than a measurement it can check.

Security features run on the same telemetry. A browser extension surfaces unapproved AI tools as employees adopt them, each detected tool is scored for data sensitivity, compliance exposure and security risk, and approval workflows let an administrator approve, monitor or block. The coding side watches agent traffic for personal data, health data and secrets, kept on a separate risk surface so the two halves do not double-count.

Olakai has opened the developer surface since the last review. A public documentation site carries a formula reference, a REST API specification and a roles-and-permissions matrix, and a TypeScript library published under an MIT licence, a Python library and a command-line tool ship through public package registries. Those libraries cover getting AI activity into Olakai. The measurement engine itself is not open to inspection, and no third-party evaluation of it appears in the reviewed sources, so the return figures a buyer would most want to check rest on Olakai's account of its own method. [s2, s3, s4, s5, s6, s7, s8, s16, s17, s18, s19]

Olakai competes on breadth and neutrality rather than depth in any one lane. Its manifesto argues that no AI vendor will build cross-vendor measurement, because doing so would mean showing how its own tool compares with a rival's. That argument holds against the AI providers and does nothing against the other neutral parties already working the same ground.

Each half of the offer meets a different specialist. DX sells measurement of AI-assisted engineering and names Adyen among the customers speaking on video. Harmonic Security sells real-time visibility and control over workforce AI across the browser and desktop, WitnessAI catalogs shadow AI applications, agents and MCP servers, and Singulr AI markets an enterprise AI and agentic control plane. Credo AI works the governance lane and displays a Forrester Wave leader placement, a form of third-party recognition Olakai's record does not carry.

The bundling pressure comes from the platforms Olakai measures. Microsoft ships its own Copilot dashboards, and Olakai's integrations page sells against exactly that, offering the visibility it says Microsoft's own dashboard will not give. The same page names Microsoft as a partner, so the platform Olakai measures is at once a channel and a plausible source of a good-enough native alternative. [s12, s26, s28, s29, s27, s30, s10]

Olakai has moved from a quote-only motion to a published price list. Starter is free for up to four seats, Olakai Assistive lists at five dollars a month an employee, Olakai Agentic at twenty-five a developer, and an Enterprise tier is quoted for regulated and large organizations that want private cloud, on-premises or managed hosting. Publishing the numbers puts Olakai in front of a self-serve buyer without waiting for a sales conversation.

Named references have appeared on Olakai's partner page. The partner page presents two founding partners with executives quoted by name, the consultancy AI Aspire through managing director Kirsty Tan and the workforce-safety vendor SolusGuard through chief executive Serese Selanders, who says the analytics turn a client pilot into a company-wide rollout. AI Aspire is disclosed on the same page as backed by AI Fund, which also backs Olakai, so one of the two references sits inside the investor's network.

The rest of the traction record is still vendor-voiced. The homepage carries four testimonials attributed to roles rather than companies, the newsroom reads that industry updates and announcements are coming soon, and no named customer, case study or independent review appears in any reviewed source. The partner program itself is concrete, publishing a referral fee band for advisors and a deployment margin band for certified firms, which is the kind of detail a consultancy needs before it builds a practice on a young vendor. [s13, s14, s1, s32, s24]

Xavier Casanova's record is the team's documented asset. AI Fund's article records his chief executive role at NASDAQ-listed Presto, his founding of Liveclicker, sold to Campaign Monitor in 2018, and his founding of Fireclick, sold to Digital River in 2004, and credits him with five patents in video and predictive technology. HP Tech Ventures counts Olakai as his sixth venture since he started Fireclick as a Stanford student. The businesses the cited record describes are a restaurant automation company and an email personalization platform, both outside security and AI governance.

The bench under him stays thin in public. Olakai's company page names Walt Mann as co-founder and chief technology officer and Paul Brzozowski as co-founder and chief revenue officer, and the reviewed sources publish no background for either. The newsroom announces no leadership hire, and the reviewed sources establish no headcount.

The backing on the record is affiliated rather than commercial. AI Fund lists Olakai in its portfolio and describes the company in its own words, HP Tech Ventures records AI Fund as the backer, and Olakai's company page names Dean Sysman as a board member. Investor and board names of that kind open doors, and none of them stands in for an arms-length customer. [s21, s22, s12, s20, s32]

Olakai has closed the assurance gap its earlier record showed open. The trust page states that a SOC 2 Type II examination is complete for a period running from March to May 2026 against the security criteria, with the report available under a non-disclosure agreement. It also documents independent penetration testing by outside security firms on a periodic basis, AWS hosting in the us-east-1 region with isolated private networking, and a named list of infrastructure and AI subprocessors.

Deployment options now answer the residency question the single-region hosting raised. The Enterprise tier offers private cloud, on-premises or managed hosting, and the pricing page states that Olakai deploys inside the customer's own cloud by default so data does not leave the customer's control. A buyer weighing where employee prompts land has a choice the earlier record did not offer.

The product concentrates exactly the data its own marketing warns about. Olakai reads prompts to find sensitive content, so it holds the material it tells buyers to protect, and it names Anthropic, OpenAI, Google, Mistral and Perplexity among the providers that process data through API calls with no persistent storage at the provider. A prompt privacy mode that requires explicit authorization before a non-administrator can read prompt content, with access attempts logged, is the control Olakai offers against that concentration. [s11, s13, s8, s15]

Add analyzed competitors to compare them side by side with Olakai.

Olakai sells to six named industries and four executive personas at once. The site carries industry pages for technology, financial services, healthcare and life sciences, professional services, retail and manufacturing, and use-case pages for the head of AI, the CFO, the CISO and the VP of engineering. That spread is ambitious for a company whose public masthead names three founders.

The published price list narrows the target. Olakai Assistive is sold to enterprises rolling AI out across the workforce, Olakai Agentic to leaders who manage AI budget and have to prove coding tools and agents are worth the bill, and an Enterprise tier to regulated and large organizations that want private cloud, on-premises or managed hosting. Each tier states its buyer in the vendor's own words.

Demand evidence still lags the marketing. The freshest customer material remains four undated testimonials attributed to roles rather than companies, the newsroom says announcements are coming soon, and no dated, named proof point appears in the reviewed sources. For a vendor selling proof of value, that absence is what a buyer would check before signing.

The claimed problems are specific and consistent across every page. Olakai names unprovable return, volatile usage-based coding spend, idle licenses, ungoverned agents and shadow AI, and each product page describes a mechanism rather than a benefit: admin-API ingestion for spend, pull-request analysis for output, a browser extension for discovery, and content policy enforcement on prompts.

Olakai has opened enough of its method to be argued with. Its documentation states that assistive value comes from estimating the minutes of human work each interaction likely saved, and that coding value comes from the measured change in code shipped per week after adoption against a loaded engineer cost. Agent return is value created divided by execution cost, where that value is again estimated time saved. The coding product classifies pull requests as AI-assisted by reading commit co-author trailers, bot authorship and body markers. A buyer can now see what the numbers are made of, and can also see that the assistive figure is an estimate rather than an observation.

The AI advantage Olakai claims is normalization, not a model. Its differentiation is reading every vendor's usage and billing data into one comparable view, with Kai answering questions across it and showing which sources were queried and which calculations ran. No proprietary training corpus, licensed dataset or cross-customer learning loop is claimed in the reviewed sources, so the AI in the product serves analysis and presentation.

The public developer surface is new and modest. A TypeScript library published under an MIT licence, a Python library and a command-line tool ship through public package registries, and a documentation site carries a REST API specification and a KPI formula reference. Those libraries cover getting activity into Olakai. The measurement engine is not open to inspection and no third party has evaluated it in the reviewed sources, so the return figures a buyer would most want to check rest on Olakai's account of its own method.

Olakai runs a self-serve motion in front of an enterprise sale. Starter is free for up to four seats with no credit card, the two paid products carry published per-seat prices, and a work-email link opens a live environment preloaded with realistic data so an evaluator can click around before talking to anyone. For a young vendor selling measurement, showing the dashboards on realistic data before procurement starts answers the credibility problem directly.

The partner program is the year's substantive change in distribution. Olakai names two founding partners with executives quoted by name, the consultancy AI Aspire through managing director Kirsty Tan and the workforce-safety vendor SolusGuard through chief executive Serese Selanders, and it publishes three partner tracks with referral fees, deployment margins, certification training and co-branded reporting. AI Aspire is disclosed on the same page as AI Fund backed, and AI Fund backs Olakai, so one of the two named references sits inside the investor's network.

Founder-led selling still carries the motion. Casanova fronts the press the company has, a co-founder holds the chief revenue officer title, and no quota-carrying sales organization is visible in the reviewed sources. With no named customer in public evidence, the partner channel is the route by which references could arrive from outside its investor's network.

Olakai has put its prices in public, which is a change a buyer can act on. Starter is free forever for up to four seats and one product edition, Olakai Assistive lists at five dollars a month an employee, Olakai Agentic at twenty-five a month a developer, and Enterprise is quoted with volume and multi-year terms. A comparison table sets out which capabilities each tier carries, down to the seat cap and the on-premises option.

The two units of charge fit the two buyers. Pricing the workforce product per employee ties it to the population being governed, and pricing the coding product per developer ties it to the population burning tokens, so neither line rides the volatility of AI spend itself. That also makes Olakai one more per-seat license inside a stack it audits for idle licenses, a comparison its own product invites.

What a heavy Kai user costs Olakai stays outside the price list. The trust page names Anthropic, OpenAI, Google and Mistral as providers that process data through API calls, and no page in the reviewed sources states which provider answers a Kai query, whether usage limits apply, or what fair-use terms attach to the per-seat price.

Olakai delivers as software with integration-first onboarding. The integrations page says a stack connects by dropping in provider admin API keys and pointing the platform at a GitHub or Bitbucket organization, with data flowing within minutes and no SDK rollout or per-tool instrumentation. The trust page details what runs underneath: containerized applications on AWS in the us-east-1 region, PostgreSQL in a private subnet, and encrypted storage.

The browser extension is the operationally heavy exception. Discovery of unapproved tools and prompt-level controls depend on deploying and maintaining an extension across the employee fleet, which is endpoint management work the minutes-to-connect claim does not cover, and coverage stops wherever the extension does not run. Direct SDK and API calls plus Zapier and n8n hooks extend capture to sources the extension and the admin APIs miss.

Deployment choice is the year's other change. The pricing page states that Olakai deploys inside the customer's own cloud by default so data does not leave the customer's control, and the Enterprise tier offers private cloud, on-premises or managed hosting with zero Olakai access to customer data. A residency-bound buyer that the single-region description would have stopped now has a path.

Olakai has closed the assurance gap its own trust page used to advertise. The trust page states that a SOC 2 Type II examination is complete for a period running from March to May 2026 against the security criteria, with the report available under a non-disclosure agreement, and it documents periodic independent penetration testing by outside security firms, named infrastructure and AI subprocessors, role-based access control, and multi-tenant isolation enforced per query.

What remains undone is narrower than before. Olakai holds no ISO 27001 or equivalent certification in the reviewed sources, the completed examination covers the security criteria alone rather than availability or confidentiality, and the report is available only under agreement, so a buyer verifies it in procurement rather than on the page.

The product concentrates exactly the data its own marketing warns about. Olakai reads prompts to find sensitive content, so it holds the material it tells buyers to protect, and it names Anthropic, OpenAI, Google, Mistral and Perplexity among the providers that process data through API calls with no persistent storage at the provider. A prompt privacy mode requiring explicit authorization before a non-administrator can read prompt content, with every access attempt logged, is the control Olakai offers against that concentration, and an enterprise buyer will still test it.

Olakai calls itself a platform and the architecture partly earns the word. Three measurement modules feed one data model that Kai reasons over, so each new source of activity makes the conversational layer more useful, and the platform page states that both products run on the same data foundation with one set of controls.

The ecosystem around it still runs one way. Olakai consumes data from many vendors through admin APIs, version control and a browser extension, and it now publishes a REST API and client libraries so a customer can send its own activity in. Nothing in the reviewed sources shows a third party building a product on Olakai, and no marketplace listing appears. The Microsoft relationship is the one named alliance, and it doubles as a dependency on a vendor with its own dashboard ambitions.

Becoming the system of record for AI decisions is the platform claim that would matter. Retained policy-acceptance records, budget definitions and return baselines are the assets that could make Olakai the place where an enterprise decides which AI to keep, cut or block. Olakai claims that position and the reviewed sources carry no customer or third party confirming it.

The founder is a commercial operator rather than a domain insider. AI Fund's article records Xavier Casanova as chief executive of NASDAQ-listed Presto, founder of Liveclicker, sold to Campaign Monitor in 2018, and founder of Fireclick, sold to Digital River in 2004, and credits him with five patents in video and predictive technology. HP Tech Ventures counts Olakai as his sixth venture since he started Fireclick as a Stanford student. Selling measurement to enterprises sits inside the analytics and marketing-technology history the cited record describes, a restaurant automation company and an email personalization platform, and that record evidences no comparable background in selling governance to a security buyer.

The bench under him stays thin in public. Olakai's company page names Walt Mann as co-founder and chief technology officer and Paul Brzozowski as co-founder and chief revenue officer, and the reviewed sources publish no background for either. The newsroom announces no leadership hire, and the reviewed sources establish no headcount.

The backing on the record is affiliated rather than commercial. AI Fund lists Olakai in its portfolio, HP Tech Ventures records AI Fund as the backer, the company page names Dean Sysman as a board member, and the partner page discloses AI Fund behind one of the two named partners. Those relationships supply introductions and validation from parties with a stake in the outcome, and the cited record carries no arms-length customer alongside them.