guava
Screening
Candidates ranked against the requirement, with the reasons shown.
Faster screening, and fewer fake candidates.
For owner-led tech staffing and recruiting firms of roughly $5M–$50M in revenue. Your recruiters get AI working on screening, job descriptions, and submittals, with flags for the candidate fraud and AI-assisted interviews now reaching every desk.
Engagements // Fixed feesLength
- [ 1 ] AI Assessment Start here $4,0002 weeks
- [ 2 ] AI Rollout 3–6 weeks
- [ 3 ] Implementation & Offer Launch 3 months
The Assessment’s fee is credited in full toward your next project.
- Fixed fee, agreed in writing
- Romeo does the work, start to finish
- You own the accounts, code, and licenses
Works with
Microsoft 365 Copilot
Custom builds on
Azure AI Foundry
Amazon Bedrock
Observability, SIEM & DLP
MSP platforms
I help your team roll out Microsoft 365 Copilot, Claude Enterprise, and ChatGPT Enterprise, and build custom AI on Azure AI Foundry, Amazon Bedrock, or Vertex AI when a workflow needs more. I tell you honestly which fits. No reseller margin, no referral fees.
- 15+ years of engineering, hands-on at companies like Airbnb, Uber & Facebook
- 5 years building AI agents and workflow automations for tech services firms, at places like Google, Pinterest & Cisco
- Background in solutions architecture, DevOps & cloud, so I know how tech services firms actually run
- Independent No vendor partnerships, resold licenses, or referral fees
- What I’ve built Two private AI benchmarks, Main Street Bench for small businesses and MSP Bench for managed IT providers, and Guava Mode, the open-source AI setup I run Guava on.
- Track record Most recently, as Chief AI Officer, I took an SF-based tech staffing company from placements to AI services. See the numbers
Where AI pays off for tech staffing firms.
Recruiter hours go to reading, writing, and checking. AI helps with all three; recruiters decide.
Job descriptions
Descriptions and outreach drafted in your voice.
Submittals
Write-ups for account managers, drafted from the résumé and the screen.
Candidate fraud
Flags for mismatched histories, identity inconsistencies, and AI-assisted interview answers, for a recruiter to check.
These are typical starting points. The AI Assessment ranks which are worth building for your firm.
From placements to AI services, and 2.4x the valuation.
As Chief AI Officer at an SF-based tech staffing company, from February to September 2026, I helped it sell AI services to the clients it already staffed. Until then, all its revenue came from placements.
- 2.4x the acquisition price a private equity firm estimated, against the year before I joined
- 6 engineers hired full time and trained, each delivering 3–4 projects
- 2 in 3 Assessments that became larger implementations
- 80% of those implementations stayed on for ongoing support
What I set up
- An in-house delivery team, where before it only matched profiles and placed people
- Offer tiers: an Assessment first, then an implementation, then ongoing support
- The funnel, website, and marketing to sell them
- Bi-weekly hackathons where clients brought an idea and saw a working prototype
- A managed services team that looks after every project once it’s live
In production for its clients
- A sales prep agent for a large tech company: account research in, branded battlecards and pitches out
- A Salesforce and finance agent for a social media company, answering routine seller questions from runbooks and past cases
- A chatbot over a finance firm’s private Snowflake and SharePoint data, with evals on answer quality
- Assistant agents for every employee in Slack, Teams, and WhatsApp
- Client onboarding for a wealth manager, started from Salesforce, with a person reviewing before anything is submitted
- One gateway for every model call at a large accounting and audit firm, routing each by cost and need
- Code-review agents in GitHub Enterprise for a large networking company
Why the valuation moved
- Revenue went from 100% placements to projects plus recurring support
- More revenue per employee
- Lower churn
- Less depending on any one person
Three ways to work together.
Most staffing firms start with an Assessment, then an AI Rollout for the recruiting desk. Fixed, published fees.
Three engagementsFixed fee · agreed in writing
01 · Start here · Fits tech staffing firms
AI Assessment
$4,0002 weeks
For when AI could help, but you can’t name the job yet.
- A review of your workflows and data
- A ranked list of where AI saves the most time, with costs
- An honest pick of platform for each team: Copilot, Claude, ChatGPT, or Gemini
- A clear yes or no on what’s worth building
02 · Fits tech staffing firms
AI Rollout
Data readiness built in
$25,0003–6 weeks
$21,000 after your Assessment credit
For a new rollout of Copilot, Claude, or ChatGPT, or for licenses you already pay for that your team barely uses.
- Your data and permissions ready first, on any platform
- 3–5 use cases with measured baselines
- Configuration and connectors
- Hands-on training for your recruiters and account managers
- Adoption tracking
Hours back per recruiter per week, measured in your ATS.
How the Rollout runs Book a 20-minute call03
Implementation & Offer Launch
MSP to MIP in 90 days
$35,0003 months
$31,000 after your Assessment credit
From managed service provider to Managed Intelligence Provider (MIP), with your own team doing the work.
- Working AI use cases inside your systems, tested against criteria agreed up front
- Custom builds on Azure AI Foundry, Amazon Bedrock, or Vertex AI where the chat tools fall short
- Your team learns to deliver AI services across Copilot, Claude, and ChatGPT
- Your AI services, packaged and priced for your clients
- By the end, your team runs it without me
AI Rollout, with your data ready first.
Three phases3–6 weeks · $25,000
Ready
Your data, before AI sees it
- Microsoft 365 tenant or Claude and ChatGPT workspace review
- Oversharing and permissions cleanup plan, from SharePoint to connectors
- Sensitivity labels and data protection basics
- Pilot group selection
Activate
On Copilot, Claude, or ChatGPT
- 3–5 use cases, each with a measured baseline
- Configuration and connectors
Adopt
Your team, on its own work
- Hands-on training on real tasks
- Adoption tracking against the baselines
Done the AI Assessment? Ready starts from its ranked list. Starting with the Rollout? Ready picks your first use cases too.
Afterwards: Ongoing Optimization & Governance, month to month.
Ongoing Optimization & Governance
A system nobody maintains quietly stops being used, and AI nobody governs spreads on its own. I coach your team to run both.
- Adoption & optimization
- Onboarding new staff, workflow updates, and usage reviews.
- Workflow operations
- Monitoring, fixes, and connector upkeep.
- Governance & security
- An AI acceptable-use policy, access and connector reviews (Claude and ChatGPT connectors, Copilot agents, admin rights), and incident guidance.
- Shadow AI
- An inventory of the AI tools, connectors, and agents in use, and a quarterly governance review.
- Run it yourselves
- Your team learns to run governance itself.
$4,000 / month
Cancel anytime.
What an AI Assessment looks like.
The deliverable itself, for a fictional firm: every workflow ranked, each with a verdict. Some say no.
Fictional firms and illustrative numbers. See all three samples on one page.
AI Assessment · Ranked opportunities
Example staffing firm 45 internal staff · ~$30M revenue · fictional
| # | Workflow | Hours back / week | AI tools / month | Verdict |
|---|---|---|---|---|
| 1 | Screening candidates against the requirement | 30 | $400 | Build |
| 2 | Submittal write-ups for account managers | 16 | $150 | Build |
| 3 | Flags for candidate fraud and AI-assisted interviews | 10 | $300 | Pilot |
| 4 | Job descriptions and outreach drafts | 11 | $120 | Pilot |
| 5 | Auto-rejecting candidatesWhy not: screening errors are hard to spot. AI ranks; recruiters decide. | – | – | Don’t build |
Next step: an AI Rollout for #1 and #2, about 46 hours a week back for about $550 a month in AI tools.
Every engagement runs the same way.
I lead at the start. By the end, you accept the work and own it.
Who leads each stepSix steps
Guava
You
01Scope
02Access
03Build
04Acceptance
05Ownership
06Afterwards
How you’ll know it workedIn every scope
-
Real tasks
Tested on work from your live queue.
-
Named limits
What AI must refuse, written down, with human review where mistakes are costly.
-
Your dashboards
Results tracked in your own account. See mine, live
Agreed in writing before work starts.
Consulting is broken. I plan to fix it.
The old wayWhat you get instead
Not Billed by the hourA fixed fee, in writingNot A strategy deckWorking systems, tested on real tasksNot A junior benchThe founder does the workNot Done when we say soDone when you say soNot Lock-inYou own everything
As Guava grows, I’ll add senior engineers who work with you directly. More about me
Fair questions.
Anything else? Ask it on the call.
Is Guava new?
The name is; the practice isn’t. Guava is the new name for my five years of building AI agents and workflow automations for clients, and it lets me add senior engineers as I grow. Most recently, as Chief AI Officer, I took an SF-based tech staffing company from placements to AI services.
See the track record See Main Street Bench See MSP Bench See my AI usage, live See Guava Mode
Which staffing workflows do you start with?
Usually screening candidates against the requirement, submittal write-ups for account managers, and job descriptions and outreach drafts. The Assessment ranks them for your desk.
Can AI help spot candidate fraud?
It can flag what deserves a second look: work histories that don’t line up, identity and location inconsistencies, and interview answers that read as if an AI assistant wrote them. A recruiter checks every flag, and nothing is rejected automatically.
Will AI make hiring decisions?
No. AI ranks and drafts; recruiters decide. I don’t recommend auto-rejecting candidates.
Why not just buy the team an AI subscription?
Microsoft 365 Copilot, Claude Enterprise, and ChatGPT Enterprise are good tools, and buying licenses is the easy part. Licenses without use cases become shelfware: people try the chat window, the work runs the same as before, and the seats renew anyway. I pick the use cases and measure where they start, get your data ready, train your team on its own work, and track adoption, so the licenses earn their cost. Already paying for seats nobody uses? The Rollout starts from the licenses you have.
Copilot, Claude, or ChatGPT: which should we use?
Whichever fits each team, and it’s often more than one. Copilot fits work that lives in Outlook, Teams, and SharePoint; Claude Enterprise and ChatGPT Enterprise often fit writing, analysis, and work across other tools. On Google Workspace, I roll out Gemini the same way. When a workflow needs a custom build, I build it on Azure AI Foundry, Amazon Bedrock, or Vertex AI, usually whichever cloud you already use. I don’t resell licenses or take referral fees, so the recommendation is only about fit.
Will you deliver AI services to our clients for us?
No. I don’t white-label or work under your brand. I get your own team delivering AI: I assess, roll out, and coach until your people run it. For MSPs, Implementation & Offer Launch is how your team learns to package, price, and deliver AI services to your clients.
Do you only work with tech services firms?
That’s my focus: owner-led and mid-market MSPs, IT services firms, and tech staffing firms, roughly $5M–$50M in revenue. Larger firms and other owner-led businesses are welcome to start with an Assessment.
What will AI cost us each month after launch?
It depends on the platform, the number of seats, and how much your team uses it. Every scope estimates it up front.
What happens to our data?
It stays in your accounts, your tenant, and your cloud. I use the narrowest access that works, log what I touch, and return admin at handover. Security and governance are part of the work: on Microsoft 365 the Rollout fixes oversharing and sets sensitivity labels before Copilot can see anything, and Ongoing Optimization & Governance keeps an inventory of the AI tools, connectors, and agents your team uses.
Do we need the Assessment before the Rollout?
No. If you’re not sure what’s worth doing or which platform fits, start with the two-week Assessment. If you already know you want Copilot, Claude, or ChatGPT, start with the Rollout. Either way nothing is done twice: the Assessment decides what’s worth doing, and the Rollout’s first phase, Ready, gets your systems ready to switch AI on safely. After an Assessment, Ready starts from its ranked list and the $4,000 is credited in full, so the Rollout is $21,000. Without one, Ready also picks your first 3–5 use cases.
How soon can we start?
An Assessment usually starts within 1–2 weeks of signing.
Bring your bottleneck.
Tell me what’s slowing your team down. I’ll tell you where to start, or that you don’t need me yet.
Book a 20-minute call with Romeo
You’ll talk with the person who does the work. Prefer email? romeo@workwithguava.com
Book a 20-minute call: workwithguava.com/talk · romeo@workwithguava.com
On the call20 minutes
- 01
I ask about your data, your volume, and who does the work today.
- 02
You leave with a recommended engagement and a fixed fee, in writing.
