The first version of Wizeline sold a tidy promise: stop guessing what customers want. Its dashboard collected ideas from across a company, ranked features by market opportunity and gave product managers a roadmap they could defend. In 2014, that sounded like salvation for teams still negotiating priorities through spreadsheets, slide decks and whichever executive spoke last.
Bismarck Lepe launched the company just after building Ooyala, the online-video platform later acquired by Telstra for roughly $400 million. He knew the mess personally. Products absorb money long before anybody learns whether users care. Wizeline's early software tried to make those bets visible. Beta customers included Chartbeat, Fairfax Media and SnappyTV, and the startup had about 16 people split across San Francisco and Guadalajara.
Then the more interesting company appeared. By 2015, Wizeline was expanding beyond SaaS into product-development services. The official history calls it a move past the limitations of software. Put less politely: a compass is useful, but many companies were also missing the crew, the engine and someone willing to cross the ocean with them.
Three versions of the same question
Customers wanted a crew, not a compass
That change rebuilt the economics. Wizeline became a B2B services business: enterprise clients buy scoped projects or longer-running teams spanning product strategy, experience design, software engineering, cloud, data, quality and operations. There is no one-click rate card because the unit is an engagement, not a seat license. The cost depends on the system being changed, the specialists required and how long Wizeline stays embedded.
The market is crowded. Accenture and Cognizant bring enormous benches. Globant, Thoughtworks, Endava and EPAM sell their own versions of digital engineering. Clients can also hire internally. Wizeline's wedge is a cross-border delivery model planted in Guadalajara and San Francisco from day one, paired with a product culture rather than a body-shop pitch. Teams are meant to fit inside the client's operation, make decisions with it and leave behind capability, not merely code.
Bismarck Lepe built software to cure roadmap guesswork, then discovered the diagnosis was bigger than the dashboard.
Capital paid for the expansion. Publicly documented rounds include $11.1 million in 2014, a $6.3 million follow-on in 2015 and a $43 million Series B led by Apax Digital in 2018. At that Series B, Wizeline was nearing 500 employees and said revenue had grown more than 200 percent year over year. CDPQ bought a majority stake in 2021, with Apax remaining a shareholder. A $20 million MidCap Financial credit line followed in 2024.
The petabyte test
Consulting claims get foggy fast, so consider the Etsy job. Its on-premises Vertica analytics warehouse had reached roughly 450 terabytes. Around it sat more than 10,000 tables, 1,000 data sources, 3,000 pipelines, 2,000 views, 1,000 SQL scripts and ten internal systems that needed refactoring. The hardware was approaching end of life. The license bill constrained growth. And turning analytics off while the team moved everything was not an option.
Fifteen Wizeline specialists worked alongside Etsy's engineers to inventory dependencies, re-architect the platform on Google Cloud, move rollups to BigQuery and validate the migration across teams. The warehouse kept running. It later grew from 750 terabytes to 1.5 petabytes, while the new setup reduced total cost of ownership by more than 30 percent. Wizeline says data engineers gained more than 40 percent more time for new models and insights instead of routine upkeep.
Other work shows the range. MiSalud used an integrated design-and-engineering studio to turn a culturally specific telehealth idea into a downloadable app in five months. Arc XP, the publishing platform born inside The Washington Post, has used Wizeline for core platform development and product evolution; Wizeline also helped onboard four media companies. These are systems where workflow, trust and accumulated context matter as much as model performance.
“The human is (still) the adult in the room.”Wizeline's useful rule for agentic delivery
A consultancy tries to acquire a memory
The AI chapter is another version of the original pivot. Clients do not need another clever demo; they need AI that survives security review, connects to real data and changes an operating metric. Wizeline first framed its answer as AI.R+: WizeAcademy for learning, WizeBrain for shared knowledge and WizeCores for reusable components. In 2025 it added Vector, an internal agentic platform that turns project evidence into risk signals and recommendations.
The company now wraps those pieces in Perform ^ AI. Purpose-built applications and agents sit inside human-led, cross-functional pods. WizeStudios create the working environment. The idea is to make a services firm behave a little more like software: learn once, reuse carefully, improve the method. Vector was running across more than a dozen client engagements in early 2025. Wize Media Suite applies the approach to broadcast and publishing, using multimodal search, smart clipping, localization and vertical-video production.
For a media company, that can mean asking for every moment in an archive where a certain player scores, finding the clips by image, sound and transcript, then reframing them for a phone screen. Wizeline's 2026 NAB demonstration used TwelveLabs video models with Amazon Bedrock AgentCore and S3 Vectors to search material without first pouring every asset into one new repository. The output can travel as an MP4 with subtitle files or move into social publishing workflows. This is less glamorous than generating a movie from a sentence. It is also much closer to an invoiceable business problem.
In finance, the pitch shifts to document intelligence, compliant content operations and workflows where every automated step needs a trail. In healthcare, it becomes triage, scheduling, patient communication and clinical summaries, with clinicians and care navigators still holding authority. Retail work spans forecasting, supply chains, personalization and customer service. Underneath those industry wrappers sit the older muscles: API design, cloud migration, data pipelines, DevOps, quality engineering and product management. Applied AI is the new storefront; software engineering remains the stockroom.
That distinction explains where Wizeline sits in the market. It is not selling a foundation model against OpenAI, Anthropic or Google. It is not primarily a packaged-software vendor asking a client to adopt one rigid workflow. It is the integration and product layer between powerful general technology and a company's peculiar systems. The promise is that domain-shaped components shorten the blank-page phase while engineers still adapt the last mile. Buyers are not purchasing intelligence by the token. They are purchasing a team that can make intelligence behave inside an enterprise.
The claimed outcomes are pointed enough to audit. Wizeline says a content-optimization pod lifted organic traffic 40 percent across priority categories for one publisher in 60 days; a care-coordination pod cut no-shows 22 percent at a regional health system; and a revenue pod improved close rates 27 percent while shortening a SaaS client's sales cycle 15 percent. Those customers are unnamed, so treat the numbers as company-reported evidence, not a universal law.
The earlier MiSalud project offers a more visible picture of the method. The team began with a ten-week prototype phase, tested build-versus-buy choices, mapped the service as well as the screens and put an alpha in front of patients and physicians. Feedback shaped the on-platform beta. The lesson is easy to borrow and hard to fake: compress the first learning loop, but do not confuse a fast prototype with a finished service. Speed came from narrowing the question and placing researchers, designers and engineers together, not from skipping the unphotogenic work.
The part worth copying
Start with a measurable business problem. Put product, engineering and domain people in one small unit. Give agents narrow jobs and access to approved data. Reuse components only after they have earned trust. Keep one human accountable for the decision. Measure the operating result, not the number of prompts sent.
Good machinery still needs good ground
This model works when a client has an owner, a metric, usable data and enough organizational patience to change a workflow. It weakens when the mandate is “add AI,” access controls are unresolved, data is unreliable or nobody can decide whether a pilot should graduate. Reusable cores speed up familiar patterns; they do not erase industry regulation, legacy architecture or internal politics. Nearshore teams also need overlapping hours, strong documentation and decision-makers who show up.
Wizeline has tried to make learning part of the infrastructure. Its Academy began offering free technical courses in 2017; a 2020 report counted 250 courses, 30,000 applicants and 20,000 students across 56 countries. A 2026 partnership with DataCamp extended role-specific AI training beyond engineers to finance, people, delivery, marketing and operations. That matters because an AI-native company made entirely of AI specialists would be a contradiction. The actual work crosses payroll, procurement, legal review, customer support and the dozens of ordinary places where software meets a business.
Leadership has changed with the thesis. Andrés Angelani became CEO in 2024 as Lepe moved to Executive Chairman. In May 2026, Inés Casares took the chief executive role after nearly two years overseeing services, AI strategy, solutions, practices and workforce. It is an operator's promotion at an operator's company. Wizeline's disclosed Q1 2025 numbers - $34 million in bookings, $32 million in revenue, seven new logos and 156 hires - provide a baseline for judging what comes next.
The through-line is almost funny. Wizeline started by telling product managers to use evidence before committing resources. Twelve years later, it is telling chief executives the same thing about AI. The dashboard disappeared, but the argument survived: decide what matters, build around the outcome and check whether anybody's life actually improved.