From 2022–25, the best software consultancies quietly re-wrote their own delivery playbooks with AI: better discovery, faster prototyping, and more honest scoping.
By 2025 almost every consultancy slide deck mentioned AI. But only some firms were using it to fundamentally improve how they worked. The rest treated AI as a new logo to bolt onto old ways of delivering software.
Where AI Actually Helped Delivery
- Discovery: summarising existing systems and documents into clear maps for stakeholders
- Prototyping: exploring multiple UX flows faster with AI-assisted design and copy
- Scoping: using historical case data to sanity-check timelines and risks
- Support: building small, focused internal tools for debugging and triaging incidents
The consultancies that earned repeat AI work were the ones who could point to how AI changed their own process - not just their customers’ slides.
From Slideware to Changed Delivery Playbooks
Firms that really internalised AI started by turning it inward. Before proposing AI roadmaps to clients, they quietly rebuilt parts of their own discovery, estimation, and delivery processes. That meant running AI-assisted discovery on their own legacy case studies, using copilots for internal tooling, and standardising on a few proven architectural patterns instead of reinventing them for every proposal.
- Discovery: using RAG assistants to map existing client systems and documents into a single, navigable brief for the team
- Estimation: comparing proposed scopes against a library of past projects with similar characteristics to sanity-check timelines
- Delivery: reusing hardened, observable agent and RAG patterns instead of ad-hoc prompt scripts per client
- Support: building internal consoles for tracing, debugging, and cost analysis that every project could lean on
New Roles and Skills Inside Consultancies
The organisational changes were just as important as the technical ones. Successful consultancies carved out clear roles for AI leads, evaluation owners, and delivery engineers who specialised in productionising prototypes. These roles sat alongside traditional product and engineering leads, not beneath them, and were involved from pre-sales through to post-launch support.
We saw a recurring pattern by 2025: firms that invested in one or two deeply skilled AI delivery leads - people who could talk to both CTOs and line-of-business owners - closed more realistic deals and delivered them more reliably than firms that tried to spin up a large but shallow 'AI practice' on day one.
How Proposals Changed When AI Was Real
Proposal decks from high-maturity consultancies looked different by late 2024. They contained fewer generic AI capability slides and more concrete, low-level plans: which systems would be integrated, how evaluation would be run, what the ingestion pipeline would look like, and which risks might force a change of approach. Evaluation, governance, and change management appeared as explicit line items with time and budget, not hidden inside a vague 'contingency' bucket.
A useful client question
When a consultancy pitches an AI project, ask: 'Can you show us how you use AI in your own delivery process today?' Firms that can only answer in theory are still learning. Firms that can show screenshots of internal tools and metrics are far more likely to ship something durable.
Building something in this space?
We'd be happy to talk through your use case. No pitch - just an honest conversation about what's feasible.
Book a 30-minute callKey takeaways
- AI-augmented discovery surfaced risks and opportunities earlier in projects
- Reusable agent and RAG patterns replaced one-off experiments
- Evaluation and governance became visible line items in proposals
- Clients increasingly judged consultancies on post-launch support, not just build speed
- The best firms used AI internally before selling it to clients