
Insights
Article 8 of 14 – State of the Market 2026
Nine ways to get software in 2026. What does each quietly leave for you to do?
Read Time: 6 min

€1,000 in subscriptions or €450,000 in day rates over six months. Nine routes, nine criteria, and the two columns we lose.
Getting the same regulated application built can start at €1,000 in tool subscriptions or run to €450,000 in day rates for a contract team, over the same six months. Those two figures aren't buying the same thing: much of the gap is work the cheaper routes hand back to you once they're done. Below are nine ways to get software (eight ways to build it, and buying a product instead), the nine criteria we ranked them against, and the two columns where SevTech come out worst.
Take a mid-sized regulated application: about forty features, three integrations, an audit trail, and accessibility requirements. What does it cost to get it built in Ireland or the UK this year? Over a six-month build, app-generator subscriptions come to €1,000–5,000, which buys you the tool, not the finished application. A five-person Dublin contract team for the same six months comes to €250,000–400,000, at €395–675 a day depending on the role. In the UK it's €300,000–450,000, with the median software developer contract at £529 a day. Those are staffed-delivery estimates, not quotes. At the two ends that's a gap of more than 400 to one between the cheapest way in and the dearest, and the way we read it, each price is buying a very different amount of finished work.
The nine routes, and what each is built for
There are nine ways to get software now: eight ways to build it, and buying a packaged product instead. Each one is set up for something different, and there's also a layer of verification tools you can add to most of them.
Augmented in-house teams. Copilot, Cursor, Claude Code, Kiro, and similar tools, at roughly $10–200 a seat a month plus usage. Built to make your existing developers faster.
Self-serve app generators. Lovable, Replit, Bolt, Base44, and Emergent, at $16–100 a month. Built to get you to a demo quickly.
Autonomous agent vendors and software factories. Devin, Factory, Blitzy, and Codex cloud. You give them a task or a specification and get back a pull request or a whole build. Pricing runs from usage credits up to Blitzy's $250,000 structured pilots.
Hyperscaler modernisation platforms. AWS Transform, GitHub's modernisation agents, Google's Dual Run, and Moderne. Built for migration, and the output points at the vendor's cloud.
Low-code and platform-native builders. Power Apps, OutSystems, Mendix, and ServiceNow's Build Agent. Built around a governed runtime, which is also where your application then has to live. Gartner expects the low-code market to reach $58.2 billion by 2029.
Services firms. From the global integrators to Irish and UK mid-tier firms such as Version 1 and Kainos. Built for accountability and for having people on the ground.
Do-it-yourself pipelines on the agent SDKs. Full control, and you build and maintain everything yourself. MIT's NANDA study found internal AI builds succeed only a third as often as tools bought from specialist vendors or built with a partner.
Verification-first autonomous delivery. A provider runs the agents for you and hands back a working application with executed test and compliance evidence, sold as an engagement with no self-serve subscription. It's the newest and smallest route. Specialist firms offer it, SevTech among them, and the global integrators are building towards it through their partnerships with Cognition (more on that below). Built for regulated builds and modernisation, where the evidence matters as much as the code.
Buying a packaged product. Where a product already does the job, you buy it instead of building anything. Built for speed: you adapt your process to the product.
Verification overlays. CodeRabbit, Qodo, Snyk, QA Wolf, and Vanta add review, scanning, testing, and governance on top of any of the above, at about $20–300 a developer a month. They don't build anything themselves, so they don't count as a route or get a row in the matrix below. That the category sells as well as it does tells you how much review and evidence the other routes leave for you to add.
Two things have changed in the last year. First, distribution has moved towards the model labs and GitHub. GitHub's Agent HQ now runs Claude and Codex agents side by side inside the repository, and Cognizant and Infosys have both built Devin into how they deliver. Second, the vendors have been buying each other. Base44 went to Wix, Windsurf to Cognition, Ergo to Presidio, CreateFuture to Version 1, Ona to OpenAI, and Tabnine to Tricentis: six deals in about fourteen months. Whichever route you pick, it's worth assuming the name on the contract could change before the contract ends.

Figure 1: Comparison of nine approaches to obtaining software, covering eight build options and buying an existing product. The figure highlights the different levels of cost, control and ongoing work involved, including what each approach leaves for the organisation to manage.
What each route leaves for you
No route comes out best on every axis. We scored all nine routes above, and scored the verification-first row on our own model, since that's the one we know in detail. The scores are our assessment from public documentation, pricing, incident records, and analyst commentary, and the method note is published with the matrix. We scored them on the nine things that decide procurement in a regulated organisation: speed to verified software, depth of verification, compliance evidence, brownfield capability, ownership and portability, EU data residency, burden on your own team, entry price, and vendor scale. The biggest difference between them is what they hand back to you when they're finished.
App generators. Fastest to a demo and cheapest to start, and they produce no compliance evidence of their own. After the €5,000 you still owe the security review, the test evidence, the data-protection controls, the accessibility audit, and the documentation. How much that costs depends on what you already have in-house, but it's a good part of what a contract team would have invoiced you for.
Assistants. Your developers get faster, and verification stays exactly where it was, with your developers.
Agent vendors. You get autonomy and full ownership of the output. Compliance evidence, independent review, and EU hosting terms are left for you to negotiate. We checked the leading developer agent: it has no first-party EU data residency, and its hosted cloud sessions need a direct subscription, so they can't run through the hyperscalers' European regions.
Hyperscaler platforms. Strong on brownfield and scale, and you stay inside the vendor's cloud.
Low-code. Strong on governance and weakest on ownership and brownfield. The application lives in the vendor's runtime, and the cost of leaving is the part that rarely gets priced.
Services firms. Score well on compliance, residency, and burden, and worst on price. The accelerator IP usually stays with the provider.
Do-it-yourself. Complete control and the heaviest burden.
Buying a product. Fastest where a product exists, and it scores lowest on brownfield and ownership, because you adapt to the product rather than the other way round.
That leaves the row we need to be most honest about: verification-first autonomous delivery, sold as an engagement with no self-serve subscription. It scores full marks on seven of the nine columns and lowest of all on the other two, entry price and vendor scale. You can't try it for $25 a month, and the firms offering it aren't hyperscalers. If those two columns matter most to you, one of the other routes is the right answer, and the matrix will show you which one.

Figure 2: Comparison of nine software delivery approaches across multiple evaluation criteria.
How to choose
Weight the columns for your own situation before you look at a price. The cheap routes are cheap in the columns a regulated buyer cares least about, and expensive in the ones they care most about. Here are five rules we'd suggest.
Price what you still have to add. Ask every route what you'll still have to add, and for the cheap routes that's most of the work. Add security testing, compliance documentation, an accessibility audit, and environments to the €5,000 subscription and the gap to the contract team it looked like it was undercutting gets a lot narrower. How much narrower depends on your own team; the point is to price it before you choose, not after.
Pay for speed to verified software. In a regulated organisation, only software with the test and security evidence behind it gets to production, so that's the speed worth paying for.
Check where the code and the model run before you check the price. Most routes only partly meet EU residency, and for frontier models the in-region route today is through the hyperscalers' European regions. An EU endpoint keeps inference in region; it doesn't settle where logs, telemetry, support access, and subprocessors sit, so ask about those separately. It is starting to move: Microsoft says Claude in Microsoft Copilot will sit inside its EU Data Boundary by the end of 2026. That covers Microsoft's Copilot family, so for standalone coding-agent services the position hasn't changed yet.
Decide your exit before you go in. Low-code and hyperscaler routes leave the application in someone else's runtime or cloud. Agent vendors and services firms leave it in yours. By year three that difference is usually worth more than the gap in day rates.
Price the outcome. HFS surveyed 202 senior leaders at enterprises with $1 billion or more in revenue: 70% expect effort-based pricing to fall over the next two years, and 75% expect outcome-linked pricing to grow. Ask every route, services firms included, to quote for the verified application instead of the days.
None of these rules points to the same route every time. A prototype for an unregulated audience belongs on an app generator. A mainframe estate belongs with a modernisation specialist. A strong team with a long backlog is well served by assistants and an overlay or two. Where it goes wrong is when a cheap route gets picked without anyone pricing what it leaves behind.
What we hold ourselves to
The verification-first row is the one we're in, and we scored it on our own model. We've published the matrix with the two columns we lose scored the same way as everyone else's. Our position is this. On the seven columns that decide regulated procurement (verified speed, verification depth, compliance evidence, brownfield, ownership, residency, and burden) we designed our delivery model to score full marks. On entry price and vendor scale, a 15-person Dublin firm can't compete, and we don't try to. That leads to three commitments.
Portable by default. Everything we build on is standard and portable, so vendor scale matters less than it looks. If we disappeared tomorrow, a client would keep a repository, a pipeline, executed test evidence, and infrastructure definitions that any of the other routes could pick up.
Residency and independence built in. We run frontier models through EU-region hyperscaler endpoints, so inference stays in region, and we document where logs and evidence are stored rather than leaving it to contract wording. Every design and every release is also reviewed by a second model from a different vendor. That doesn't replace independent assurance, but it means a single vendor's blind spot can't pass unchallenged.
Priced on the verified application. We quote for the working, verified application and leave day rates out of it. That's what Forge was built to make possible, and article eleven in this series goes into it.
So a question for your own organisation. For the last piece of software you commissioned, which of the nine columns did you actually weight, and which did you find out afterwards you should have?
Sources:
1. NoCode.MBA, Lovable pricing explained, updated 30 October 2026. https://www.nocode.mba/articles/lovable-pricing
2. Morgan McKinley, 2026 salary guide Ireland, technology contract rates, 2026. https://www.morganmckinley.com/ie/salary-guide/technology/contract-salaries
3. ITJobsWatch, UK software developer contract day rates, six months to 19 November 2026. https://www.itjobswatch.co.uk/contracts/uk/software%20developer.do
4. Hauer Power, Nearshore software development rates 2026, April 2026. https://www.hauerpower.com/en/insights-posts/nearshore-software-development-rates-2026
5. GitHub, GitHub Copilot plans, September 2026. https://github.com/features/copilot/plans
6. GitHub, GitHub Copilot is moving to usage-based billing, April 2026. https://github.blog/news-insights/company-news/github-copilot-is-moving-to-usage-based-billing/
7. Blitzy, Product and pricing pages, September 2026. https://blitzy.com/
8. Cognition, Devin pricing, September 2026. https://devin.ai/pricing
9. Factory, Pricing and Series C announcement, April–September 2026. https://factory.ai/pricing
10. Anthropic, Claude plans and API pricing, September 2026. https://claude.com/pricing
11. Anthropic, Claude Code feature availability by provider (documentation), November 2026. https://code.claude.com/docs/en/feature-availability
12. Gartner, Forecast analysis: low-code development technologies, worldwide, 4 November 2025. https://www.gartner.com/en/documents/7146430
13. Fortune, reporting MIT NANDA, The GenAI Divide: vendor and partnership builds succeed about 67% of the time, internal builds a third as often, 18 August 2025. https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/
14. CodeRabbit, Pricing, September 2026. https://www.coderabbit.ai/pricing
15. GitHub, Pick your agent: use Claude and Codex on Agent HQ, February 2026. https://github.blog/news-insights/company-news/pick-your-agent-use-claude-and-codex-on-agent-hq/
16. Cognizant, Cognizant and Cognition partner to scale autonomous software engineering, January 2026. https://investors.cognizant.com/news-and-events/news/news-details/2026/Cognizant-and-Cognition-Partner-to-Scale-Autonomous-Software-Engineering-and-Deliver-Business-Value-Across-Enterprise-Operations/default.aspx
17. Infosys, Infosys and Cognition collaborate to accelerate the AI value journey, January 2026. https://www.infosys.com/newsroom/press-releases/2026/collaboration-accelerate-ai-value-journey.html
18. CNBC, Cognition valued at $10.2 billion two months after Windsurf deal, 8 September 2025. https://www.cnbc.com/2025/09/08/cognition-valued-at-10point2-billion-two-months-after-windsurf-.html
19. Ona (formerly Gitpod), Ona is joining OpenAI, June 2026. https://ona.com/blog
20. Tabnine, Tabnine joins Tricentis, July 2026. https://www.tabnine.com/blog/
21. Wix.com, Wix further expands into vibe coding with acquisition of Base44, June 2025. https://www.wix.com/press-room/home/post/wix-further-expands-into-vibe-coding-with-acquisition-of-base44-a-hyper-growth-startup-that-simplif
22. Silicon Republic, Dublin's Version 1 to acquire CreateFuture, May 2026. https://www.siliconrepublic.com/business/dublins-version-1-to-acquire-createfuture-ireland-uk-ai
23. Presidio, Presidio signs definitive agreement to acquire Ergo, October 2025. https://www.presidio.com/news/presidio-signs-definitive-agreement-to-acquire-ergo/
24. HFS Research, The headcount era is fading: the pivot to outcome and consumption pricing, July 2026. https://www.hfsresearch.com/research/headcount-era-fading-pivot/
25. Keyhole Software, How much does custom software development cost in 2026, January 2026 (published build-cost benchmarks). https://keyholesoftware.com/cost-custom-software-development/
26. Sacra, Blitzy company profile (pilot and contract pricing), November 2026. https://sacra.com/c/blitzy
27. Microsoft, Bringing Claude models in Microsoft Copilot into the EU Data Boundary, 5 November 2026. https://www.microsoft.com/en-us/copilot/blog/2026/11/05/bringing-claude-models-in-microsoft-copilot-into-the-eu-data-boundary/
Made with AI, not by it. Claude researched and drafted; I rewrote and verified it. The charts are Claude’s.
View more Insights from SevTech.
News

SevTech at Web Summit 2026 with Enterprise Ireland.

Jean Gargan-Smith
Marketing & Operations Lead
State of the Market Series

Your 2027 compliance deadlines are a software delivery problem

Colm Lucey
Director & Senior Information Technology Consultant
State of the Market Series

Where AI-built software actually pays for itself. So why are the demos all somewhere else?

Paul Mahon
Director & Senior Technology Consultant
State of the Market Series

One in five Irish enterprises now uses AI. So why is the average the least useful number?

Fionnuala Healy
Director of Innovation & AI Strategy
State of the Market Series

The analysts are undercounting it. So what is the forecast measuring?

Rob Curley
Managing Director
State of the Market Series

The €96 billion question. Why is it really a €710 billion one?

Daithí Thornton
Sales Director
State of the Market Series

‘Passing’ is a measurement, not a claim. So who ran the test?

Hugh Nolan
Director & Senior Architect
State of the Market Series

One in three pull requests now involves an agent. So why does nobody trust the code?

Fionnuala Healy
Director of Innovation & AI Strategy
News

SevTech is growing, bringing new expertise and experience to our team.

Jean Gargan-Smith
Marketing & Operations Lead
News

SevTech’s Summer Internship Programme

Rob Curley
Managing Director