Andrea Sarmiento · Technical Product & Specification Manager · Dubai
Hi, I'm Andrea.
Imakecomplexproductseasytounderstand,buyanduse.
I'm a Technical Product & Specification Manager with an engineering background. I've built CRMs, product portfolios and AI workflows across construction, industrial and e-commerce, always with the same method: start from the business metric, find the friction, ship in phases.
An engineer's rigour. A product manager's perspective.
"A product only matters if someone can use it on Monday morning."
01 · Roots
From Colombia to construction sites
I studied civil engineering, so my first products were buildings. On site and in consulting for oil & gas and real estate, I learned how products really get chosen, installed and used.
02 · The switch
Into the commercial side of products
In Dubai I moved into real estate and CRM, then coatings, and now construction chemicals at Pidilite Puma MEA. That is where product management found me: deciding what to build, for whom, and proving it with numbers.
03 · My edge
I speak engineer and I speak business
I connect engineering, R&D, sales and marketing: I can write a CSI specification, give a CPD seminar, model a price in SAP, then design the CRM stage and the AI workflow that track it.
04 · Next
The method travels
Construction is where I learned the craft, not where it ends. I want to bring the same approach to new products and industries, from industrial to e-commerce and tech.
How I work
1
Start from the business metric
Pipeline, conversion or hours saved: I decide what has to move before I choose a tool.
2
Find the friction with the people who live it
I sit with sales, specifiers and managers to see where work gets lost or repeated.
3
Ship in phases and measure
I launch a small first version with real users, then build only what the data supports.
Tell me what you are launching, fixing or scaling. I'm open to product management roles in tech, industrial, e-commerce and B2B, remote or flexible. I'm based in Dubai and work in English and Spanish.
Projects lived in personal spreadsheets and were forgotten during a 1 to 2 year sales cycle.
What I did
Proposed and rolled out a CRM: clean data, stages designed with sales, live market data and 1:1 adoption.
Result
15 users in 4 teams track an AED 9.5M pipeline in one dashboard.
Case 1 of 3
01Context
Dubai has a project on every corner. We were losing them in spreadsheets.
When I joined, every team worked its own way: personal Excel files, notes and inboxes. We would visit a project and then lose track of it, and by the time someone followed up, a competitor was already in. Reports took days, and managers could not see how many projects we had or who owned each one.
Click through the buying journey the CRM was built around
02Challenge & goal
I proposed one system for the whole company, built to outlast any one of us.
The goal: one place for every account, project and conversation, with one owner per client and a real pipeline value, so the work stays even when someone leaves. HubSpot and Salesforce were over budget, so I chose CorporateStack, a mid-market CRM already used by companies like ours and known to some of my colleagues.
Specific
One CRM for every account, project and conversation, with one clear owner per client.
Measurable
A real pipeline value, stage by stage, that management can check at any time.
Achievable
Within budget: no HubSpot or Salesforce, a mid-market CRM the team could adopt.
Relevant
Our sale closes 1 to 2 years after specification. A forgotten project is a lost sale.
03My role
What I owned, end to end.
1
Proposed the CRM and chose the platform within budget
2
Cleaned and migrated ~2,000 records with AI
3
Designed stages, flags and probabilities with the sales team
4
Ran a 4-person pilot and set up branded quotation templates
5
Integrated BNC and Apollo for live project data
6
Trained every user 1:1 and now manage the admin panel
Construction needs a CRM built for a two-year sales cycle.
1-2 years between specification and purchase
I had used CRMs before. I first built a small version in Notion, but it could not track the values we needed. Other construction-material manufacturers used the same CRM, with one account per salesperson. Our difference: we sell finishing materials, so the purchase comes 1 to 2 years after the specification, at the very end of the project. Without a system, a project specified today is forgotten by the time it is ready to buy.
2
Phase 2 · Clean the data
I cleaned the data before building anything.
~2,000 records in one clean database
I gathered every Excel file and used AI to map them into the CRM import template, removing duplicates along the way. Around 2,000 records went in: accounts, projects, opportunities and sales, assigned to their owners one by one.
3
Phase 3 · Design the stages
I designed the stages with the people who sell.
I sat with the sales team to agree on the stages that matter and the flags inside each one: approved, on hold, not approved and resubmit, plus a Hot flag for priority projects. The default stage probabilities did not fit construction, so we rewrote them.
4
Phase 4 · Pilot
We piloted with four people before the rollout.
4 people in the pilot, then 4 teams
Four of us used it first and fixed what did not work: branded quotation templates, one template per client type, email signatures and fields nobody used. The CRM provider built each change we agreed on. Today I manage the admin panel, and any new field or flag is agreed in a meeting first.
5
Phase 5 · Connect live data
I connected live market data instead of chasing it.
BNC + Apollo feeding the CRM in real time
I proposed adding BNC, a project-intelligence platform, and integrated it with the CRM, so a project's status updates on its own as it moves from design to tender to construction. Apollo fills the contact gaps. With both, we got specified on large projects where we were not on the original list.
6
Phase 6 · Win adoption
I won adoption one person at a time.
1:1 training until everyone was on board
The hardest part was people, not software. Some colleagues were not digital natives, so I ran inductions and one-to-one sessions to show how the CRM saves them time. When salespeople disagreed over shared clients, we added a contribution field to split commissions, and we are now moving to one owner per account.
05Key decisions
The calls I made, and what they cost.
01
CorporateStack over HubSpot or Salesforce
Why
The budget ruled out the big names, and companies like ours already used it, so colleagues knew it.
Trade-off
Fewer features out of the box, so we rewrote the stage probabilities for construction.
02
Clean the data before launch
Why
A CRM full of duplicates loses people’s trust on day one.
Trade-off
A slower start: around 2,000 records mapped and assigned before anyone logged in.
03
Pilot with 4 people first
Why
Real use shows what breaks: templates, signatures, fields nobody needs.
Trade-off
The full rollout came later, but it came with fewer complaints.
06Impact
The whole pipeline, in one dashboard.
AED 9.5Mlive pipeline, tracked in one dashboard
Company revenue · AED
7MBefore
15MThat year
CRM users
Pilot · 4 peopleToday · 15 users in 4 teams
15 users across 4 teams track an AED 9.5M pipeline. The dashboard shows how long deals take between stages, value lost each month, who is selling and where deals stall. Close dates tell us when to follow up. Company revenue grew from AED 7M to 15M that year, and the CRM gave sales the pipeline visibility behind that growth.
TodayThe team runs its pipeline in the CRM, and we are moving to one owner per account.
07Lessons
Looking back.
What I learned
Adoption is the product. The software was the easy part; the people were the project.
What I would do differently
Agree on one owner per account from day one, instead of fixing shared clients later.
Let’s talk about a project like this.
I make complex products easy to understand, buy and use.
Localising a European product portfolio for the GCC luxury market
How do you get an unknown brand specified in a market where price often beats quality?
Company Pidilite Puma MEAWhere GCCWhen 2024 - presentRole Senior Technical Product & Specification ManagerTools SAP · CSI MasterFormat · TDS
2 / 2years with sales targets met
Full systemsfor every application scenario
Problem
An unknown European brand entering a GCC market that buys on price.
What I did
Repositioned the pitch, rebuilt our documents for the GCC, and sold complete systems and a value line.
Result
Sales targets met in year 1 and 2, and specified on projects such as La Palma and DAMAC Lagoons.
Case 2 of 3
01Context
A European range, launched into a market that had not asked for it.
The joint venture brought Grupo Puma's cementitious range to a market that judges products by GCC standards, price and local wording. The plan was to lead with tile adhesives, our core expertise. The problem: architects are not excited by adhesives, and nobody knew the brand.
02Challenge & goal
My goal: get on the specification first. The sale comes later.
My job: put the brand on consultants' specifications and adapt the range to GCC buyers, with a small team and a product line still being built locally.
Specific
Put the brand on consultants’ specifications and make the range fit GCC buyers.
Measurable
Hit the yearly sales target, starting in year one.
Achievable
With a small team, samples from Spain and a local line still in production.
Relevant
Finishing materials are specified 1 to 2 years before purchase. Being specified first means selling later.
03My role
What I owned, end to end.
1
Redesigned the pitch for CPD and Lunch & Learn sessions
2
Restructured the company profile, catalogues and data sheets for the GCC
3
Supported marketing: event sponsorships and exhibitions such as The Big 5
4
Put together complete systems for each application
5
Built competitor comparisons from public TDS
6
Wrote CSI specifications and technical presentations
I opened with what catches the eye and closed with what we make.
I did not agree with leading with adhesives. I proposed opening Lunch & Learn and CPD sessions with eye-catching products from Spain, like decorative plasters, and closing with the adhesives we make here. I brought my manager to meetings to see the difference, and it became our standard pitch.
2
Phase 2 · Pre-launch
We presented products before manufacturing them.
1-2 years head start from the specification cycle
Finishing materials are specified 1 to 2 years before they are bought. So we presented products before producing them locally, using samples from Spain, and were already on approved lists when production started.
3
Phase 3 · Build visibility
We put the brand where the industry meets.
The Big 5 and sponsored industry events
Nobody specifies a brand they have never seen. Alongside the technical work, we invested in marketing for visibility: sponsoring industry events and taking part in major exhibitions such as The Big 5 in Dubai, where consultants, contractors and developers meet.
4
Phase 4 · Localise the documents
I rebuilt our product documents for the GCC market.
GCC first in standards, wording and units
A European company profile and catalogues do not sell in the Gulf as they are. I helped restructure and reconfigure the company profile, the product catalogues and the data sheets for this market. Technical values come from R&D; I changed how they read: local phrasing, units, commas for thousands and GCC standards first, because consultants here care more about GCC norms than foreign ones. Convincing Spain and India took many meetings; in the end, results won the argument, not slides.
5
Phase 5 · Sell systems
We started selling systems, not products.
With a small team we launched material by material: adhesives first, our core strength, then the products that complete each scenario. Complete systems now open doors to new segments, such as swimming pool contractors.
6
Phase 6 · Value line
We competed on value without lowering the standard.
Our comparisons showed higher quality, but also a higher price, in a market that buys on price. Some products cost too much to produce, so we launched a value line: good performance at a lower production cost. We go to the limit of the standards, never below.
05Key decisions
The calls I made, and what they cost.
01
Open with showpiece products, close with adhesives
Why
Architects get excited by decorative plasters, not by tile adhesives.
Trade-off
Our core product moved to the end of the pitch.
02
Present products before manufacturing them
Why
Finishing materials are specified 1 to 2 years before they are bought.
Trade-off
We relied on samples from Spain while local production caught up.
03
A value line, never below the standard
Why
In a market that buys on price, our quality came with a higher price.
Trade-off
A second line to manage, with the standards as the floor, never below.
06Impact
From unknown brand to specified on flagship projects.
2 / 2years with sales targets met
Sales target
Year 1✓Target met
Year 2✓Target met
Why being specified first wins
SpecifiedWe present the system
1-2 years head start
PurchasedWe are already on the approved list
We met our sales targets in year 1 and year 2. Using BNC, Apollo and AI to reach the right people, we were specified on projects such as La Palma in Jebel Ali and DAMAC Lagoons in Dubai, including some where we were not on the original list.
TodayComplete systems keep opening new segments, such as swimming pool contractors.
07Lessons
Looking back.
What I learned
Results win internal arguments, not slides. Proof from the market convinced Spain and India.
What I would do differently
Start adapting the documents for the GCC at the same time as the first presentations, not after.
Let’s talk about a project like this.
I make complex products easy to understand, buy and use.
Two AI agents that turn days of technical document work into minutes
How do you cut days of technical writing without losing accuracy?
Company Pidilite Puma MEAWhen 2026Role Agent designer and ownerTools Microsoft 365 Copilot · Claude
2 days → 1 hper CSI specification
2 days → 30 minper competitor comparison
Problem
Each CSI specification and competitor comparison took about 2 days by hand.
What I did
Built two no-code AI agents with locked rules, and dropped the ideas that did not earn their place.
Result
1 h per CSI and 30 min per comparison, my review included.
Case 3 of 3
01Context
Two days per specification, and every update meant starting again.
2 days per specification, by hand
Each CSI specification took me about 2 days, and any new product or TDS revision meant rewriting it. Competitor comparisons took another 2 days and were rebuilt from scratch for every request.
02Challenge & goal
I wanted AI in everyone's daily work, used safely.
My manager and I proposed Microsoft 365 Copilot for the whole company and organised training for every team. At the same time, I wanted AI to take over my most repetitive technical documents.
Specific
AI that drafts CSI specifications and competitor comparisons straight from the TDS.
Measurable
Time per document, down from about 2 days.
Achievable
No-code: the company Copilot licence plus Claude, no developers.
Relevant
Every new product or TDS revision meant rewriting documents by hand.
03My role
What I owned, end to end.
1
Proposed Copilot company-wide and co-led the training
I learned through YouTube, the Microsoft AI Product Manager Professional Certificate on Coursera and daily use of AI at work and at home. I started with ChatGPT, then tested tools for each job: writing, images, agents and code. Copilot is best for internal documents; Claude handles large comparison tables far better.
2
Phase 2 · Lock the rules
I locked what must never change.
The first drafts had errors, so I gave the agents rules: copy TDS values and fixed wording exactly, follow my CSI structure, and write "To verify" when data is missing. The agent drafts; I review, adjust and approve.
3
Phase 3 · Build in Copilot
I built the Specification Assistant in Copilot.
It runs in Agent Builder on the company licence, so internal documents stay inside Microsoft 365, and it writes the full three-part CSI specification from the TDS in our company format.
4
Phase 4 · Build in Claude
I built the Competitor Comparison as a Claude skill.
It keeps one living comparison of 6 brands, organised by EN classification. It answers quick questions, such as us vs. one brand, and logs every change when a TDS is revised.
5
Phase 5 · Cut what fails
I dropped what did not earn its place.
2 ideas dropped that did not earn their place
A submittal agent handled a sales task but still left the PDF assembly to do by hand, so I dropped it. A product selector for sales was too risky without an engineer checking the answers.
05Key decisions
The calls I made, and what they cost.
01
Copilot for specifications, Claude for comparisons
Why
Copilot keeps internal documents inside Microsoft 365; Claude handles large comparison tables far better.
Trade-off
Two tools to maintain instead of one.
02
Locked values and a human review
Why
The first drafts had errors, and a wrong value in a specification is expensive.
Trade-off
1 hour instead of minutes, because I check every specification.
03
Drop the submittal agent and the product selector
Why
One still left the PDF assembly by hand; the other was too risky without an engineer.
Trade-off
Fewer features, more trust in the ones we kept.
06Impact
One hour per specification, thirty minutes per comparison.
2 days → 1 hper CSI specificationReview included
CSI specification
Before
2 days
After
1 h
Competitor comparison
Before
2 days
After
30 min
CSI specifications went from about 2 days to 1 hour including my review, and competitor comparisons from 2 days to 30 minutes. My role moved from writing to technical judgement.
TodayBoth agents are part of my daily work: the agent drafts, I review and approve.
Before · by hand2 daysper CSI specification2 daysper competitor comparison
Copy every value from the TDS
Rewrite for each TDS revision
Rebuild comparison tables from scratch
After · with AI agents1 hper CSI specification, review included30 minper competitor comparison
The agent drafts from the TDS
I review, adjust and approve
Every TDS change is logged
⟷
Try it: ask the comparison skill
The AI builds the comparison from each brand's TDS. Pick a class and a competitor to see the kind of answer it gives, with real values in N/mm².
Class
Compare with
Pidilite Puma values are measured ranges from third-party test reports; competitor values are as stated in their public TDS, often declared minimums (≥).
07Lessons
Looking back.
What I learned
Each AI is good at something different. Pick the tool for the task, not the other way round.
What I would do differently
Write the locked rules before the first draft, not after the first errors.
Let’s talk about a project like this.
I make complex products easy to understand, buy and use.
