The verdict in three sentences
For a Dublin services firm sending more than 15 proposals a month, a custom AI proposal generator pays for itself in under a year. The value does not come from the language model but from the library of approved blocks (case studies, methodologies, clauses, CVs) that the tool assembles. Budget 12,000 to 30,000 EUR excluding VAT and 6 to 10 weeks as a 2026 ballpark, plus 50 to 300 EUR a month of API usage.
Why a proposal still takes 3 hours
In most consulting, engineering and IT services firms around Dublin, proposals are built by copy and paste: reuse the last winning bid, hunt for the right client reference in a shared drive, rewrite the methodology, then check pricing by hand. A sales director spends about 3 hours on a 20-page document, and much longer on a public tender submitted through eTenders with a detailed technical response.
The problem is twofold. Time first: 20 proposals a month means 60 hours, nearly two weeks of a senior manager's work. Quality second: recycled blocks carry outdated references, another client's name or stale rates. A well designed AI tool fixes both, provided it is fed with approved content.
| Proposal step | Manual time | With a custom AI tool |
|---|---|---|
| Reading the RFP and extracting requirements | 40 min | 5 min (automatic summary) |
| Picking relevant client references | 25 min | 1 min (semantic search) |
| Writing methodology and timeline | 60 min | 3 min draft, 20 min review |
| Pricing and rate card | 30 min | 1 min (pricing rules) |
| Layout and final proofreading | 25 min | 10 min |
| Total | 3 h | about 40 min |
What a serious proposal generator contains
An effective tool rests on four components. The block library: around 200 items approved by management (company overview, 40 case studies, 30 methodologies, anonymised CVs, legal clauses, terms and conditions). The semantic search engine, which finds the blocks closest to the client's stated need. The language model, which adapts wording to context without inventing facts, under a strict instruction to quote existing blocks only. Finally, Word or PDF export in the company's branding, with version tracking and manager sign-off.
CRM integration (HubSpot, Pipedrive, Salesforce) prefills the client, contact and deal value, then logs the sent proposal against the opportunity. That connection is often what separates a gadget from a tool people actually use.
| 2026 option | Upfront cost | Recurring cost | Main limitation |
|---|---|---|---|
| ChatGPT or Copilot used manually | 0 to 30 EUR per user | 20 to 30 EUR per seat per month | No approved library, risk of invented facts |
| Proposal SaaS (Qwilr, PandaDoc type) | 0 EUR | 35 to 70 EUR per seat per month | Weak on long technical tender responses |
| Simple custom tool | 12,000 to 18,000 EUR | 50 to 150 EUR a month of API | One document type only |
| Full custom tool with CRM | 18,000 to 30,000 EUR | 150 to 300 EUR a month of API | Requires real content work |
| Maintenance and upgrades | quoted separately | per contract | Block updates to plan for |
These are 2026 ballpark figures from the Irish and wider EU market. The most underestimated item is building the library: plan 3 to 5 days of internal time to sort and approve the 200 blocks.
A typical 6 to 10 week timeline
Weeks 1 and 2: audit of the last 30 proposals, won and lost, and splitting them into blocks. Weeks 3 to 5: building the search engine, templates and export. Weeks 6 and 7: CRM connection and pricing rules. Weeks 8 to 10: testing on live bids, tuning instructions, training the team. Data stays hosted in the EU and confidential blocks are not used to train the model, which you should require in writing under GDPR.
Mini case study
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
Aoife, sales director of a 45-person engineering consultancy in Dublin, sends 20 proposals a month. Before: 20 x 3 h = 60 hours. After: 20 x 40 min = about 13 hours. Monthly saving: 47 hours. At a fully loaded cost of 75 EUR per hour, that is 3,525 EUR a month, or 42,300 EUR a year. With a 22,000 EUR tool and 200 EUR a month of API (2,400 EUR a year), payback comes in roughly 7 months.
The commercial effect matters more. Because Aoife now answers 30 tenders instead of 20, with better targeted submissions, her positive response rate on tenders rises 1.6 times over six months. With an average contract of 25,000 EUR, two extra wins a quarter fund the tool several times over.
FAQ
Will the AI invent client references?
Yes, if left unconstrained. A serious tool restricts the model to the 200 approved blocks and flags any sentence without a source. A 10 to 20 minute human review remains mandatory.
Does our sales data go to OpenAI or elsewhere?
Business APIs offer EU hosting and no training on your data. The contract must state it, and the most sensitive documents can run on an EU-hosted model for an extra 100 to 300 EUR a month.
What does the API cost day to day?
For 20 to 40 proposals a month, usage sits between 50 and 300 EUR a month depending on the model and document length. That is marginal next to 47 hours saved.
Do we need to change CRM?
No. The tool connects to HubSpot, Pipedrive or Salesforce through their APIs. The connection usually takes 1 to 2 weeks of the project.
Who maintains the block library?
A sales lead, 2 to 4 hours a month. A simple admin screen lets them add a case study or retire an outdated block without a developer.
Let's scope your project. Send us three recent proposals and your current CRM, and we will scope the block library, integrations and a 6 to 10 week plan. Detailed quote within 48 h. WhatsApp +221 77 596 93 33.
Mohamed Bah
Fondateur, Kolonell
Passionate about digital and entrepreneurship in Africa, Mohamed has been helping Sénégalese businesses with their digital transformation since 2020. Founder of Kolonell, he believes every SME deserves a professional and accessible online présence.
