# Boutique AI consultancies in Europe, 2026

> Nine independent AI consultancies in Europe compared: base, best fit, where each is the wrong choice and how an engagement runs. Reviewed November 2026.

Author: Sara Maia (Partner & COO, Twistag), written with Claude · Category: AI & Agents · Published: 2026-11-01 · Updated: 2026-11-01

_Reviewed 1 November 2026. Twistag wrote this page and Twistag is on it. The same criteria were applied to every firm, each section says which problems that firm is the wrong choice for, Twistag included, and every claim links to its source. Nobody paid to appear._

A boutique AI consultancy is the right call when the problem is well defined, the people who scope the work need to be the people who build it, and the company intends to own the result. Which boutique depends on the problem. Twistag fits a mid-market company that needs an AI pilot taken into production or an AI-native product built by embedded senior engineers. DareData fits an Iberian enterprise buying a programme on a hyperscaler stack. Alice Labs fits a Nordic mid-sized company that wants a fixed price on a single use case. ML6 fits a Benelux or DACH enterprise that wants vision, voice or robotics work with a large partnered bench. Xomnia fits a Dutch organisation whose first job is a data platform. dida fits a German public body or industrial company with a research-grade modelling question. deepsense.ai fits a company that needs GPU infrastructure, evaluation and agent engineering with direct lines to the model labs. Augusta Labs fits a private-equity portfolio company that wants embedded transformation pods. Mantis NLP fits a foundation or grant-maker with a document-heavy NLP problem and a team to upskill.

The reason the match matters is the failure rate. [Gartner expects over 40% of agentic AI projects to be cancelled by the end of 2027](https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027), on cost, unclear value or weak risk controls. [ISG's 2025 survey of 400 senior AI decision-makers found that 31% of enterprises deploying AI have at least one of their top three funded use cases in production](https://www.businesswire.com/news/home/20250915236466/en/ai-use-cases-double-though-business-outcomes-lag-ambition-isg-study). A partner whose published work describes systems that are running is the strongest signal available before a contract is signed.

## What counts as boutique here

Four tests, applied to every firm.

**Independent.** The firm is run by its founders or partners and is neither a subsidiary of a larger group nor a private-equity roll-up. Minority stakes are allowed where the firm discloses them, and this page says so.

**Fewer than 150 people.** Large enough to staff a programme, small enough that senior people are on the work.

**AI delivery is the core business.** Firms that sell software outsourcing with an AI label, and product companies that cannot be hired, are excluded.

**Verifiable.** Each firm was checked on its own website on 1 November 2026. Where a figure comes from a directory or the press instead, the link says so.

Three well-known names were considered and left out on the first test. [Faculty is wholly owned by Accenture](https://faculty.ai/en-gb/insights/news/faculty-welcomes-halfspace-in-next-phase-of-global-growth), following the [$1bn deal announced in January 2026](https://faculty.ai/en-gb/insights/news). [Tooploox became a Solvd Inc. company in June 2025](https://tooploox.com/tooploox-joins-solvd-accelerating-ai-innovation-for-the-intelligence-age). [Datatonic is backed by the private-equity investor Perwyn and acquired Syntio in April 2025](https://datatonic.com/insights/acquisition-syntio-expansion/). All three are credible partners; they are a different kind of purchase from the firms below.

## The comparison

| Firm | Base | Best for | Where it is the wrong fit | Typical engagement |
|---|---|---|---|---|
| Twistag | Lisbon | Taking a pilot into production; AI-native products; embedded senior engineers for a mid-market roadmap | Open research questions; programmes that need a Big Four logo; pure capacity against a finished backlog | [Readiness sprint, fixed-scope project, or FDE pods of 1–3 engineers in 13-week quarters](https://twistag.com/forward-deployed-engineering) |
| DareData | Lisbon | Enterprise AI programmes in Iberia on Google Cloud, Microsoft or NVIDIA, with an agent platform | Buyers who want no platform dependency; work far from the Iberian reference base | [Discovery, then an AI MVP, then the GenOS platform](https://www.daredata.ai/) |
| Alice Labs | Stockholm | Nordic mid-sized companies buying a single use case at a fixed price with EU AI Act compliance built in | Multi-system enterprise integration; buyers who need named references before signing | [Fixed-price projects from €80,000; 4-week discovery sprint €25,000–40,000](https://alicelabs.ai/en/ai-implementation-partner) |
| Augusta Labs | Lisbon | Private-equity portfolio companies and enterprises wanting embedded pods and a client-owned AI layer | Buyers who need published case studies; small scopes | [6–8 week value foundation, gate, then 6–12 month transformation](https://www.augustalabs.ai/) |
| ML6 | Ghent | Benelux and DACH enterprises wanting vision, voice, robotics or agent programmes with a partnered bench | Companies that want a small team and no platform; the largest firm on this page | [Advisory, engineering projects and the Unum platform; 11 weeks average time-to-value](https://www.ml6.eu/en/ml6-explained/about/history) |
| Xomnia | Amsterdam | Dutch enterprises and public bodies whose first job is a cloud data platform on Azure, Databricks or AWS | Work outside the Netherlands; AI-native product work | [Consultancy, moving to multi-year projects after the 2026 merger](https://xomnia.com/?p=2216) |
| dida | Berlin | German and Austrian public sector and industry with research-grade computer vision or NLP problems | Buyers who need quantified business outcomes as evidence; product interfaces | [Feasibility study 2–3 months, production 3–9 months, then maintenance](https://dida.do/machine-learning-solutions) |
| deepsense.ai | Warsaw and Palo Alto | Agent engineering, evaluation, MLOps and GPU infrastructure with OpenAI, Anthropic and Google Cloud partner status | Customer-facing products with their own interface; buyers wanting a local Western European team | [Discovery, then advisory, a production implementation or an embedded pod](https://deepsense.ai/ai-implementation-partner/) |
| Mantis NLP | London, Valencia, Nicosia | Foundations, grant-makers and global-health bodies with document-heavy NLP work and teams to upskill | Anything needing a bench beyond a handful of people | [45-minute scan, discovery, then iterative delivery](https://www.mantisnlp.com/) |

Team sizes, where published, are in each firm's section. Several firms publish none; that is noted.

## Twistag — applied AI company and AI agency, Lisbon

[Twistag is an applied AI company founded in Lisbon in 2016](https://twistag.com/about), with [25+ engineers, designers, data and AI specialists averaging eight or more years of experience](https://twistag.com/about). It serves mid-market companies through three practices: [AI & Agents, Data & Cloud and Product Engineering](https://twistag.com/), with [clients in Europe, North America and Asia](https://twistag.com/about). It runs an [ISO 27001 practice, is an AWS partner, appeared in the Deloitte Technology Fast 500 EMEA in 2023 and 2024, and holds a Clutch rating of 4.9 from 14 verified reviews](https://twistag.com/about).

The work that defines it is integration and product. For a [European hospitality group with 2,000+ employees, a unified data layer across property management systems and an optimisation engine cut agency staffing spend by 70% in four months](https://twistag.com/case-studies/ai-workforce-optimization-hospitality-group-europe). For a [Portuguese construction company, 250+ site workers now run on one operations platform](https://twistag.com/case-studies/construction-operations-platform-portuguese-construction-company). [Twenty-two case studies are published](https://twistag.com/case-studies).

**How an engagement runs.** Three entry points: an [AI Readiness Sprint, a fixed-scope project, or Forward Deployed Engineering](https://twistag.com/about). In the FDE model, [pods of one to three senior engineers work inside the client's own repository, CI, backlog and cloud; the first quarter is 13 weeks with weekly reporting against a week-one baseline, pricing is per engineer-day billed monthly, and the client keeps the code, specs, agents and dashboards](https://twistag.com/forward-deployed-engineering). Twistag also [works through consulting and private-equity partners](https://twistag.com/#ways-to-work).

**Where Twistag is the wrong fit.** When the question is whether the thing is predictable at all, applied research comes before engineering, and dida or deepsense.ai have deeper benches for it. When the risk being managed is political, some boards need a Big Four name on the paperwork and a 25-person firm cannot supply it. When the architecture is decided and the need is engineers against a written backlog, that is staff augmentation, and paying for Twistag's judgement is paying for something already decided.

## DareData — enterprise AI on a platform, Lisbon

[DareData is a Lisbon-based, remote-first firm](https://www.daredata.ai/) that [grew past 130 AI and data specialists in 2025 after more than 60 hires, and opened an office in Alvalade](https://www.daredata.ai/resources/newsroom/artificial-intelligence-made-in-portugal-we-grew-more-than-60-in-one-of-the-most-competitive-markets-in-the-world-says-daredata-partner). The site does not state a founding year; [Clutch lists 2019](https://clutch.co/profile/daredata). It is a [Google Cloud partner, Microsoft partner and NVIDIA preferred partner, and was in the Deloitte Technology Fast 50 Portugal in 2025](https://www.daredata.ai/). [NOS, Portugal's largest telecommunications operator, acquired 20% of DareData's capital in 2024](https://www.daredata.ai/resources/newsroom/artificial-intelligence-made-in-portugal-we-grew-more-than-60-in-one-of-the-most-competitive-markets-in-the-world-says-daredata-partner) and is also its most documented client.

[Eighteen success stories are published](https://www.daredata.ai/success-stories). The named ones include [NOS invoice automation, around 20,000 invoices a month with roughly 65% processed straight through, delivered in six months on the GenOS supervisor](https://www.daredata.ai/success-stories/invoice-processing-automation-for-telco-operations), and [simulation-driven cable design for COFICAB with more than 50,000 designs tested virtually](https://www.daredata.ai/success-stories/simulation-driven-r-d-for-cable-design). The logo wall includes [Heineken, Coca-Cola, Roche, EDP, Euronext, Worten and CTT](https://www.daredata.ai/).

**How an engagement runs.** [Discovery and assessment, then an AI MVP, then custom engineering and the GenOS platform](https://www.daredata.ai/). The [homepage offers an AI MVP in four to eight weeks](https://www.daredata.ai/); the [offer page itself describes two to three months to production and says over 95% of its projects reach production](https://www.daredata.ai/solutions/ai-mvp). Worth asking which applies. [Clutch lists a minimum project of $50,000](https://clutch.co/profile/daredata).

**When it is the better fit.** A board-sponsored programme in Portugal or Spain where procurement wants a firm that already sells to companies of that size and holds the hyperscaler partnerships.

**When it is the wrong fit.** A buyer who wants no platform in the relationship. [GenOS and its applications are now the centre of the offer](https://www.daredata.ai/genos), so the contract should say what happens to the system if the relationship ends. The published references are overwhelmingly Iberian.

## Alice Labs — fixed-price AI for the Nordics, Stockholm

[Alice Labs AB was founded in Stockholm in 2023](https://alicelabs.ai/en/about) by [Linus Ingemarsson and Eric Lundberg](https://alicelabs.ai/en/about). It sells [AI training, consulting, automation, agents, search and AI SEO](https://alicelabs.ai/en/) to [finance, healthcare, supply chain and professional services](https://alicelabs.ai/en/) across [Sweden, the Nordics and the EU](https://alicelabs.ai/en/about), with [EU AI Act and GDPR compliance presented as built in from day one](https://alicelabs.ai/en/about).

The clearest thing about Alice Labs is its commercial transparency. [Fixed-price commitments, with overruns absorbed by the firm; single use cases from €80,000; a four-week discovery sprint at €25,000–40,000; retainers from €15,000 a month; eight to sixteen weeks to production](https://alicelabs.ai/en/ai-implementation-partner). Few firms on this page publish anything comparable.

The evidence is thinner than the numbers. The site claims [100+ production AI implementations since 2023](https://alicelabs.ai/en/about) and [85% of clients seeing ROI within 12 months](https://alicelabs.ai/en/ai-implementation-partner), without a list or a method. The [case page shows six cases](https://alicelabs.ai/en/case), one named: [an AI agent for the vertical farm Ljusgårda, with $250,000 a year saved and an 83% cost reduction, built in six weeks](https://alicelabs.ai/en/case/ljusgarda-ai-agent). [Reference customers are available under NDA](https://alicelabs.ai/en/ai-implementation-partner). Team size is unstated; [six people are pictured](https://alicelabs.ai/en/ai-agency).

**When it is the better fit.** A Nordic mid-sized company with one well-bounded use case, a fixed budget and a compliance officer who wants EU AI Act language in the statement of work.

**When it is the wrong fit.** Multi-system enterprise integration, or any buyer who needs named references and a disclosed team before signing.

## Augusta Labs — transformation pods for portfolio companies, Lisbon

[Augusta Labs describes itself as an applied AI lab founded in Lisbon in 2024](https://www.augustalabs.ai/), selling [AI transformation through hybrid pods of two to four people, embedded engineers who hand over to the client's team, and "The Kernel", a client-owned, model-agnostic AI layer running in the client's own cloud](https://www.augustalabs.ai/). The homepage says [60+ AI engineers](https://www.augustalabs.ai/); the [talent page says 50+ people](https://www.augustalabs.ai/talent). [Sage announced a global partnership in January 2026, with Augusta Labs engineers embedded in Sage's product organisation](https://www.sage.com/investors/investor-downloads/press-releases/2026/01/sage-partners-with-augusta-labs-to-accelerate-the-build-out-of-its-ai-centre-of-excellence/). The firm [announced a funding round in June 2026](https://www.augustalabs.ai/#news); [ECO reported a €50 million valuation with the amount undisclosed](https://eco.sapo.pt/2026/06/02/augusta-labs-fecha-ronda-com-unicornios-nacionais-e-chuta-avaliacao-para-50-milhoes/).

**How an engagement runs.** [An AI Value Foundation of six to eight weeks targeting two or three use cases in production, a one-to-two-week evaluation gate, then a six-to-twelve-month transformation and a handoff](https://www.augustalabs.ai/).

**When it is the better fit.** A private-equity portfolio company or large enterprise that wants a pod inside the business and an AI layer it owns at the end. The site cites [partner PE firms with $1.5 trillion in combined assets under management](https://www.augustalabs.ai/).

**When it is the wrong fit.** A buyer who needs published case studies: [Sage is the only named client](https://www.augustalabs.ai/), the other logos are unlabelled and no outcomes are published. Small, single-system scopes are below the engagement model.

## ML6 — the large boutique, Ghent

[ML6 was founded in 2013](https://www.ml6.eu/en/ml6-explained/about/history) and has [140+ people across Ghent, Berlin, Munich, Amsterdam and Eindhoven](https://www.ml6.eu/en/ml6-explained/careers/team). It reports [150+ clients, 400+ projects and an average of 11 weeks to value](https://www.ml6.eu/en/ml6-explained/about/history). Partnerships cover [AWS, Anthropic, Cisco, ElevenLabs, Google, Microsoft, NVIDIA and OpenAI](https://www.ml6.eu/en/technologies/partners), and it became an [OpenAI services partner in June 2025](https://www.ml6.eu/en/news/ml6-becomes-openai-services-partner-to-tackle-advanced-ai-engineering). Ownership is not disclosed on the site.

Named cases include [Scout24's HeyImmo assistant, Mediahuis NewsSearch, a Genscom curator with 3x revenue, and an NCSC deployment at 85% lower implementation cost](https://www.ml6.eu/en/customers/cases/). The 2025–2026 repositioning is towards an own platform, [Unum, sold as customer-owned agent workflows with human approval](https://www.ml6.eu/en/superintelligence/unum), and intelligence modules for [vision, physical AI, voice and computer use](https://www.ml6.eu/).

**When it is the better fit.** A Benelux or DACH enterprise that needs breadth: several workstreams, hardware-adjacent AI, and a partner tier on more than one hyperscaler.

**When it is the wrong fit.** A company that wants a small senior team for one system with no platform attached. At 140+ people and five offices, ML6 sits at the edge of what this page calls boutique.

## Xomnia — data platforms first, Amsterdam

[Xomnia was established in 2013](https://xomnia.com/Careers) in Amsterdam. It [acquired Aurai in June 2025 and merged it fully on 1 July 2026, reaching 120 specialists](https://xomnia.com/?p=2216), and [appointed a new CEO, Stephan Hartgers, in April 2026](https://xomnia.com/news/). Its partners are [Databricks and Microsoft](https://xomnia.com/about-us/). Services run from [data and AI strategy through cloud data platforms on Azure, AWS and Polars to MLOps, GenAI and agentic AI](https://xomnia.com/Services). Ownership is not disclosed.

The published cases are platform-heavy and Dutch: [Alliander, a 98% cost reduction using Polars on AWS; Tata Steel predictive maintenance; KLM delay prediction; a.s.r. data platform migration; The Ocean Cleanup streaming platform](https://xomnia.com/cases).

**When it is the better fit.** A Dutch enterprise or public body whose AI ambition is blocked by its data estate, especially on Azure or Databricks.

**When it is the wrong fit.** Work outside the Netherlands, where the firm has no office, or AI-native product work with an interface people use daily. The 2026 merger and leadership change are recent; ask who specifically would be on the team.

## dida — research-grade ML, Berlin

[dida is a Berlin firm of over 30 mathematicians and physicists](https://dida.do/about-us), founder-led by [Lorenz Richter and Philipp Jackmuth](https://dida.do/about-us), selling [ML solutions, consulting, operations and research](https://dida.do/) with strengths in [LLMs, NLP and computer vision](https://dida.do/). It has a document-processing product, [smartextract](https://dida.do/software). Projects named on the site include [idealo, ESA, Enpal, Deutsche Bahn, the German weather service DWD, Ströer and Schroff/nVent](https://dida.do/projects), and public-sector work for [four German federal ministries and Bundesdruckerei](https://dida.do/public-sector). In January 2026 it signed a [four-year AI framework agreement with Land Oberösterreich](https://dida.do/news), and in October 2026 it [ranked first in the Cyberagentur HEGEMON evaluation](https://dida.do/news).

**How an engagement runs.** [A feasibility study or proof of concept, usually two to three months; production software, three to nine months; then maintenance](https://dida.do/machine-learning-solutions), plus [workshops and coaching for client data teams](https://dida.do/machine-learning-consulting).

**When it is the better fit.** A German or Austrian public body or industrial company whose problem is a hard modelling question in vision or language, with procurement that values research credentials.

**When it is the wrong fit.** A buyer who needs quantified business outcomes as evidence; [none of the published projects state one](https://dida.do/projects). A product that needs a polished interface will need a second partner.

## deepsense.ai — agents, evaluation and infrastructure, Warsaw

[deepsense.ai has 120 AI experts](https://deepsense.ai/our-credentials/), offices in [Warsaw and Palo Alto](https://deepsense.ai/about-us/), and has [provided AI solutions since 2014](https://deepsense.ai/about-us/). It is an [official OpenAI services partner and Anthropic services partner, a Google Cloud Gemini Enterprise implementation partner, and works with AWS and LangChain](https://deepsense.ai/our-credentials/). It reports [200 commercial AI projects, an NPS of 82, and 67% of revenue from relationships older than two years](https://deepsense.ai/our-credentials/). Ownership is not disclosed.

Named outcomes include [Nielsen, 93% faster data extraction; NOAA, a 98% reduction in image-analysis time; Volkswagen reinforcement learning; and SES AI, a 128-GPU infrastructure](https://deepsense.ai/case-studies/). It publishes open-source tooling such as [ragbits for agentic RAG](https://deepsense.ai/rd-hub/ragbits/).

**How an engagement runs.** [Discovery, then one of three entry points: advisory and planning, a focused production implementation, or an embedded engineering pod sized to the roadmap](https://deepsense.ai/ai-implementation-partner/).

**When it is the better fit.** A company whose hard part is model evaluation, agent infrastructure, MLOps or GPU capacity, and that values a partner with direct lines into the model labs.

**When it is the wrong fit.** A customer-facing product with its own interface; the published work centres on models, data and infrastructure. A buyer who wants a team based in Western Europe.

## Mantis NLP — the micro boutique, London, Valencia and Nicosia

[Mantis NLP was co-founded in 2021 by three former Wellcome Trust colleagues](https://www.mantisnlp.com/about-us/) and has [offices in London, Valencia and Nicosia](https://www.mantisnlp.com/about-us/). [Clutch lists two to nine employees and a $10,000 minimum project](https://clutch.co/profile/mantis-nlp). It sells [AI opportunity identification, tool selection, process redesign, upskilling, responsible-AI guidance and agentic solutions](https://www.mantisnlp.com/), and names [GSMA, Wellcome Trust, Oxford Medical Simulation and dmTech](https://www.mantisnlp.com/) among clients. The one quantified outcome is [a 35% increase in location extraction across 130,000+ Wellcome grant applications](https://mantisnlp.com/case/helping-the-wellcome-trust-track-its-global-impact/).

**How an engagement runs.** [A discovery call, a 45-minute Business Quick Scan, then iterative delivery with the stated aim of building the client's own capacity](https://www.mantisnlp.com/).

**When it is the better fit.** A foundation, grant-maker, global-health body or association with document-heavy NLP work and a team that needs to learn alongside the build.

**When it is the wrong fit.** Anything that needs a bench, a 24/7 operation or an enterprise programme. A handful of people cannot carry it, and the firm's own positioning says so.

## Which of these to call

**A pilot that works and cannot get into production.** Twistag. The failure is rarely the model; it is the data underneath and the integration around it.

**An AI-native product that people will use every day.** Twistag, on the strength of a decade of product engineering.

**A board-sponsored programme in Portugal or Spain on Google Cloud or Microsoft.** DareData.

**One use case, one fixed price, in the Nordics.** Alice Labs.

**A PE portfolio company wanting an embedded pod.** Augusta Labs, with the caveat that outcomes are not yet published.

**Vision, voice or robotics at enterprise scale in Benelux or DACH.** ML6.

**A Dutch data estate that has to be fixed first.** Xomnia.

**A research-grade modelling problem in Germany or Austria.** dida.

**Agent infrastructure, evaluation or GPU capacity.** deepsense.ai.

**A foundation or grant-maker with documents to understand.** Mantis NLP.

## Frequently asked questions

### What is a boutique AI consultancy?

An independent firm, typically under 150 people, whose core business is designing and building AI systems and whose senior people work on the engagements they sell. On this page that excludes subsidiaries of larger groups and private-equity roll-ups, which is why [Faculty](https://faculty.ai/en-gb/insights/news/faculty-welcomes-halfspace-in-next-phase-of-global-growth), [Tooploox](https://tooploox.com/tooploox-joins-solvd-accelerating-ai-innovation-for-the-intelligence-age) and [Datatonic](https://datatonic.com/insights/acquisition-syntio-expansion/) were left out.

### When does a boutique beat a large consultancy?

When the project has to survive production. A large firm brings brand assurance and a deep bench, which matters when the risk is political. A boutique brings the people who will do the work, which matters when the risk is technical. [Gartner's forecast that over 40% of agentic AI projects will be cancelled by 2027](https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027) is an argument for picking on delivery evidence.

### How much does a boutique AI engagement cost in Europe?

Only two firms on this page publish numbers. [Alice Labs prices single use cases from €80,000, discovery sprints at €25,000–40,000 and retainers from €15,000 a month](https://alicelabs.ai/en/ai-implementation-partner). [Twistag prices forward deployed engineering per engineer-day, billed monthly, in 13-week quarters](https://twistag.com/forward-deployed-engineering). Directory minimums give a floor: [DareData at $50,000](https://clutch.co/profile/daredata) and [Mantis NLP at $10,000](https://clutch.co/profile/mantis-nlp). Any firm quoting a figure before understanding the problem is guessing.

### What should a buyer check before signing with a boutique?

Four things. Whether the published work describes systems that are running, with outcomes. Who owns the code, models and infrastructure at the end; [Twistag](https://twistag.com/forward-deployed-engineering), [Augusta Labs](https://www.augustalabs.ai/) and [ML6](https://www.ml6.eu/en/superintelligence/unum) all state client ownership, and DareData's offer should be read for how [GenOS](https://www.daredata.ai/genos) is licensed. Who specifically will be on the team. And what happens if the relationship ends after a platform is in place.

### Why trust a comparison written by one of the firms on it?

It should be read with that in mind. What Twistag has done is state the bias at the top, apply the same four tests to every firm, say for each firm, itself included, which problems to take elsewhere, and link every claim to a source that can be checked. Then call two firms from this list.

## Sources

- Twistag — [twistag.com/about](https://twistag.com/about), [forward deployed engineering](https://twistag.com/forward-deployed-engineering), [case studies](https://twistag.com/case-studies), read 1 November 2026

- DareData — [daredata.ai](https://www.daredata.ai/), [success stories](https://www.daredata.ai/success-stories), [AI MVP](https://www.daredata.ai/solutions/ai-mvp), [newsroom, 10 March 2026](https://www.daredata.ai/resources/newsroom/artificial-intelligence-made-in-portugal-we-grew-more-than-60-in-one-of-the-most-competitive-markets-in-the-world-says-daredata-partner), [Clutch profile](https://clutch.co/profile/daredata)

- Alice Labs — [alicelabs.ai/en/about](https://alicelabs.ai/en/about), [implementation partner page](https://alicelabs.ai/en/ai-implementation-partner), [cases](https://alicelabs.ai/en/case)

- Augusta Labs — [augustalabs.ai](https://www.augustalabs.ai/), [Sage press release, January 2026](https://www.sage.com/investors/investor-downloads/press-releases/2026/01/sage-partners-with-augusta-labs-to-accelerate-the-build-out-of-its-ai-centre-of-excellence/), [ECO, 2 June 2026](https://eco.sapo.pt/2026/06/02/augusta-labs-fecha-ronda-com-unicornios-nacionais-e-chuta-avaliacao-para-50-milhoes/)

- ML6 — [history](https://www.ml6.eu/en/ml6-explained/about/history), [team](https://www.ml6.eu/en/ml6-explained/careers/team), [partners](https://www.ml6.eu/en/technologies/partners), [cases](https://www.ml6.eu/en/customers/cases/), [Unum](https://www.ml6.eu/en/superintelligence/unum)

- Xomnia — [merger announcement, July 2026](https://xomnia.com/?p=2216), [about](https://xomnia.com/about-us/), [cases](https://xomnia.com/cases), [careers](https://xomnia.com/Careers)

- dida — [about](https://dida.do/about-us), [projects](https://dida.do/projects), [public sector](https://dida.do/public-sector), [ML solutions](https://dida.do/machine-learning-solutions), [news](https://dida.do/news)

- deepsense.ai — [credentials](https://deepsense.ai/our-credentials/), [about](https://deepsense.ai/about-us/), [case studies](https://deepsense.ai/case-studies/), [implementation partner](https://deepsense.ai/ai-implementation-partner/)

- Mantis NLP — [mantisnlp.com](https://www.mantisnlp.com/), [about](https://www.mantisnlp.com/about-us/), [Wellcome case](https://mantisnlp.com/case/helping-the-wellcome-trust-track-its-global-impact/), [Clutch profile](https://clutch.co/profile/mantis-nlp)

- Excluded firms — [Faculty on Accenture ownership](https://faculty.ai/en-gb/insights/news/faculty-welcomes-halfspace-in-next-phase-of-global-growth), [Tooploox on joining Solvd](https://tooploox.com/tooploox-joins-solvd-accelerating-ai-innovation-for-the-intelligence-age), [Datatonic on Perwyn and Syntio](https://datatonic.com/insights/acquisition-syntio-expansion/)

- Gartner, [press release, 25 June 2025](https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027)

- ISG, [State of Enterprise AI Adoption 2025, press release 15 September 2025](https://www.businesswire.com/news/home/20250915236466/en/ai-use-cases-double-though-business-outcomes-lag-ambition-isg-study)

---

Source: https://twistag.com/thinking/boutique-ai-consultancies-europe-2026
