Invoice-PO reconciliation
+90h/month
Pay per measurable FTE hours freed. If we don't free the agreed hours, you don't pay the variable. If we exceed them, we win together.
Strategy · Processes · Applied AI
The piece that fits your business with AI.
We work with SMBs and large corporations to set the strategy, redesign the processes and ship AI that moves the needle. No fluff. Measurable results in weeks.
Spanish SMBs of 20 to 500 employees free up 15-30 hours per week of repetitive work in 4-12 weeks. No endless POCs.
"Monthly close: from 5 days to 4 hours in 8 weeks."
Sector pattern · Industrial SMB
37%
more productivity in AI-augmented roles (McKinsey, 2025)
3-5×
average ROI 12-18 months post-deployment (IDC + ONTSI 2024)
15-30h
weekly freed per team (ONTSI benchmarks + sector pilots)
0
projects without contractual success metric
What we do
We don't sell catalogs. We diagnose the missing piece and fit it.
01 / Strategy
Tech strategy
12-to-36-month roadmap focused on EBITDA, not buzzwords. Roadmap, business case, governance and dashboard.
- + AS-IS / TO-BE diagnosis
- + Business case with ROI per initiative
- + Executive governance & RACI
02 / Processes
Process redesign
We map the chaos, kill the friction, automate the boring. Lean, BPMN and some common sense.
- + Process mining + value stream
- + RPA + AI automation
- + KPIs and continuous improvement
03 / Applied AI
AI that prints money
No endless POCs. Use cases prioritized, in production, with a business metric attached from day one.
- + Vertical copilots, custom-built
- + Back-office agents
- + Integration with your stack
Sector- representative results.
Illustrative cases based on public sector projects and benchmarks from ONTSI, McKinsey and Spanish SMB RPA adoption studies. Details of our own engagements we share on the call.
Industrial · SMB 100-300 employees
F-04 · BI
−96% close time
BEFORE
Monthly close: 5 days, 3 people dedicated, frequent errors
AFTER
Monthly close: 4 hours, 1 review, live dashboard
B2B Distribution · SMB 30-80 employees
F-02 · Admin Agent
+90h monthly hours freed
BEFORE
~1,200 invoices/month processed manually by 2 people
AFTER
85-95% automated, human validation only on exceptions
Professional services · SMB 50-150 employees
F-01 · Sales Copilot
+30-50% pre-sales time freed
BEFORE
Artisan pre-sales, proposals in 4-5 days
AFTER
Proposals in 30 min, deal close prediction
Traditional consultancy vs Puzzle
What you'll notice in the first month of the engagement.
Traditional consultancy
Puzzle Tech & Sec
Diagnosis
3 months, 40-page deck
2 weeks, 1 actionable page
Time-to-production
12-18 months
6-12 weeks
Pricing model
By the hour (you pay regardless)
Pay-per-result available
Assigned team
Senior → Junior after sale
Same team start-to-finish
Deliverables
Slides + recommendation
System in production + KPIs
Lock-in
Mandatory renewal
Exit any sprint
Focus
Maximize billable hours
Maximize your EBITDA
Five operational commitments.
Not slogans. Each one has an evidence page behind it.
- 01
Puzzle Tech is the only AI consultancy in Spain that publishes 30 real SMB scenarios mapped to EU AI Act category, applicable obligations and budget range.
EU AI Act Pillar 2026-2027 · 30 scenarios - 02
We work with success metrics agreed on each project.
Services · 0 projects without an agreed metric - 03
No handoffs between sales and delivery: the technical lead who ships each project is the same one in the first client call.
About · how we work - 04
Pay-for-performance available when the business metric is attributable and auditable.
Pricing · payment models - 05
From first contact to production in 4-12 weeks vs industry average 12-18 months (Gartner 2024 for SMB AI POCs).
Methodology · 4 phases in weeks
We share risk when the metric is attributable.
No Big Four, boutique or niche AI shop offers this model. It's our real wedge against the market — and we only apply it when the business metric is clear, measurable and attributable to the piece.
Dormant account recovery
% revenue
Pay as % of incremental revenue attributable to the system. No new revenue, no variable. Cap agreed upfront for budget predictability.
Proposal generation
Win rate
Variable pay based on win-rate uplift over a 3-month baseline. Only if there's a solid measurement method and a recent, reliable baseline.
Does it apply to any project?
No. Only when the business metric is clear, measurable and attributable: FTE hours freed, error rate, account recovery, win rate. Not for knowledge management, decision improvement, onboarding.
Who bears the base cost risk?
We do, in part. The technical setup is fixed by milestones (covers real cost). The variable is outcome-based. Cap agreed upfront so your CFO has budget predictability.
How is the metric measured?
Audited baseline pre-pilot (3 historical months). Measurement system agreed by contract. Monthly report with metric and attribution, auditable by your finance team.
What if the metric isn't met?
Variable is 0. The fixed portion is still charged for work delivered. The difference: we don't pocket variable for nothing. You don't pay variable without outcome.
Let's talk · 30 min
We'll tell you on the first call whether your case fits pay-for-performance or needs a fixed-by-milestones model.
Applied AI for SMBs, no BS.
Ready-to-fit capabilities. Live in weeks, not quarters. Pay-per-result available.
F-01
Sales Copilot
Reads your CRM, preps meetings, drafts proposals and predicts which deal closes this quarter.
+90 min/week freed (Microsoft, 2025)
F-02
Admin Agent
Processes invoices, matches POs, reconciles banks. Human validation only on exceptions — the rest automated, 24/7.
85-95% invoices auto
F-03
Ops Assistant
Turns chaotic WhatsApps, emails and PDFs into structured ERP orders.
60-80% handling time cut
F-04
Business Intelligence
Ask in natural language. Auditable answers from your own data. Monthly close stops being a project.
From minutes to seconds
F-05
24/7 Customer Support
Solves 80% on first contact in your team's tone and guidelines. Escalates the rest to humans with full context.
80%+ first-contact resolution
F-06
Internal Knowledge Graph
Your corporate knowledge, indexed and alive. Onboarding from 3 months to 3 days.
Onboarding months to days
Tailored use cases
Each sector has its pieces. These are ours.
Four moves. No theater.
A method that doesn't ask for blind trust: each phase has a deliverable, a metric and an exit point.
- 01
Diagnosis
2 weeks. We map business, data and processes. We exit with a ROI-prioritized opportunity map.
DELIVERABLE · Map + business case - 02
Design
3–4 weeks. We define the solution, integrations, operating model and change plan.
DELIVERABLE · Blueprint + plan - 03
Deployment
6–12 weeks. We implement in sprints with weekly metrics. Real production, not demo.
DELIVERABLE · System in use - 04
Scale
Ongoing. We optimize, expand use cases and transfer capability to your team.
DELIVERABLE · Autonomy + KPIs
Where would you start ?
Three questions. We recommend the highest-ROI piece for your situation. Zero strings.
1 / 3
What hurts most day-to-day?
See how much AI saves you per month.
Move the 4 sliders. We'll send you a personalized report with the 3 pieces that pay off best in your case.
Estimated result
67,894 €
Savings / month
1,940h
Hours saved / month
814,733 €
Annual savings
< 4 months
Estimated payback
No spam. Sent in less than 5 minutes.
Indicative estimate based on real projects with SMBs of 20–500 employees.
Playbooks you ship on Monday.
Download them free. If they help, we'll talk later.
Guide
The Sales Copilot playbook
How to ship a copilot that reads your CRM and doubles conversion in 60 days.
→ Read now Strategy
Process map to detect AI pieces that move the needle
Operational template to map processes in 2 90-min sessions and exit with 3-5 prioritized pieces.
→ Read now Compliance
GDPR + RAG: shipping AI with sensitive data
Secure RAG architecture, self-hosted vs managed API, minimum contract clauses and final legal checklist.
→ Read nowWe don't work with everyone.
The sooner we say it, the better for both. If you don't fit these profiles, we save you the call.
We're not for you if:
- You shop on price. We're 30-40% more expensive than standard-tariff consulting.
- You want an endless POC to 'learn'. We work in production from week 6.
- Your company has no digitized data. We fix it first — but we tell you.
- There's no C-level sponsor. Without one, AI doesn't land.
- You're looking for 'magic'. AI is engineering, not snake oil.
We're a great fit if:
- You have 20–500 employees (also large corps with autonomous BUs).
- Your CEO/COO is willing to put in 2h/week during the project.
- You want results measured in EBITDA, not innovation NPS.
- You're ready to change processes, not just add tech.
- You like direct people who say 'no' when 'no' is the answer.
What everyone wants to know.
The actual objections we hear on every first call. Unfiltered.
Pricing per project, not by catalog. The only fixed price is the initial Flash Diagnosis: €1,500 (4-5 days, credited toward the next step if you proceed). Everything else is calibrated by sector + type of piece + client magnitude. You'll find 9 representative examples in /pricing with typical ranges: industrial SMB €60-130k year 1, B2B distribution €30-75k, professional services €50-95k. Payment models: fixed by milestone, pay-per-result or hybrid depending on the case.
No. By default we use enterprise APIs (OpenAI, Anthropic, Azure): your data is not used for training. If you need it, we deploy open-source LLMs in your cloud or on-premise. Security doc on the first call.
Every phase has an exit point. If after diagnosis we don't see a case, we tell you and walk. If pilot metrics aren't hit, we adjust or cancel at no extra cost. We don't sell what we don't believe.
We work with your team, not behind their back. Co-create, train, transfer capability. By project end your team operates the system without us — that's success for both.
C-level sponsor: 2h/week. Process owner: 4-6h/week in active phases. Rest of the org barely anything. If we ask for more, something's wrong.
Yes. We work with clients in regulated sectors (financial, healthcare, public sector). DPIA, processing register, processor contracts, risk assessment — all included when applicable.
Top-one issue, that's why we don't ship raw LLMs. Guardrails, citations, human-in-the-loop on critical paths, regression tests. Production cases run <2% error on real data.
Yes, from 20 employees up. Our sweet spot is 50-300. Below 20 there's usually no business case for us — we tell you and recommend tools you can buy yourself.
30 minutes. Zero sales pressure.
We tell you exactly what will happen so there are no surprises. If at the end you don't see value, you walk away — no sales scripts, no hard pitch, no aggressive follow-up.
- 5 min01
We listen
Tell us your operational situation and where it hurts today. No slides upfront, no endless discovery.
- 15 min02
Live diagnosis
We map 2-3 candidate pieces with defensible ROI criteria. We tell you which one we'd attack first and why.
- 5 min03
Honest verdict
Three possible answers: your case fits our Flash, it fits your team + a part-time advisor, or you don't need AI yet.
- 5 min04
Next steps
If it makes sense to continue, we plan. If not, we recommend what to do and who to contact. No commitment.
If at the end of the 30 minutes you don't see value in what we tell you, you walk away.
Not ready to book yet?
Every Director General arrives at a different stage. Three ways to start the conversation depending on where you are.
01
Take the deck
12 editorial slides with our thesis, services and representative cases. Ideal if you're in research mode.
See materials 02
Explore 3 playbooks
Operational guides with templates and checklists for process owners. Read them on web without registration, ship them on Monday without needing to talk to us.
See playbooks 03
Book 30 minutes
Live diagnosis with ROI criteria. No scripts, no commitment. If you don't see value, you walk away.
Let's talk · 30 min The missing piece
Let's talk for 30 minutes. No strings.
Name and email is all we need. Company and team size help us come prepared — optional. We get back within 24h with a proposed time for a 30-minute call.