Honeymoon Israel · AI plan · August 2026

We heard three things.

“I don’t think I have clarity of what the possibilities are… and what it costs and what the tradeoffs are.”Avi, our first conversation
“Adoption is the real constraint, not the tech.”Greg, workflow map, Aug 11
“There’s no AI owner yet. Whoever we assign will matter as much as the tooling.”Greg, workflow map, Aug 11
Honeymoon Israel
This deck answers all three.
The map we propose

Six workflows, one picture.

At the end of the chain: one dashboard, with per-user permissions, runs most of this map - scoring, document status, funder reports, and page numbers.

#WorkflowOwnerWhy this order
1Application review & scoringTrip operationsStart here the why: next slides
2Applicant emails & document trackingTrip operationssame team, the win compounds
3info@ responsesinfo@ staffdaily, visible to the whole org
4Grant reporting (37 funders)Developmenta second team joins
5Recruitment marketing
part of this needs no AI at all · we’ll show you where !		Laura & Jenlater in the chain
6Personalized applicant pagesMarketinglater in the chain
7Four initiatives of our ownUsbeyond the map near the end of the deck
One at a time, each proves itself - and every external touchpoint keeps human sign-offSalesforce stays the system of record
Where it stands today

The tools are in the building.
The work is still manual.

19Claude seats across the org
7staff through a 7-week training
3using it day to day, in practice
Your own read: “most people default to old habits”That’s not a people problem
The approach
“Here are the tools. Build yourself an agent.”
We build the agent around the person. Nobody has to become a builder.
The agent adapts to how your team already worksNot the other way around
Step 1

Application review & scoring

Workflow 1 · Application review & scoringThe first move

Start where you already pointed.

Highest value

  • Every application read by hand today
  • 15 trips recruiting now, 25 more planned

Cleanest to define

  • Clear inputs: applications in Salesforce
  • One catch: the criteria live in people’s heads, written nowhere. We fix that first

Most visible win

  • “Staff need to see one of these work” - as you put it
  • A ranked shortlist beats a stack of raw applications
Your map, Aug 11: “Suggested first build”Next: how it will run
Workflow 1 · Application review & scoringA working prototype already exists !

How it will run, on your systems.

HMI your part · US ours · both badges = together

Scores and reasoning will flow back to Salesforce automaticallyFinal invite decision: always yours
Workflow 1 · Application review & scoringThe four-week plan

How we build it, week by week.

The full schedule - a part that finishes early rolls straight into the next link.

Week 1
HMIUS

Rubric, wire & build

One working session - an hour or two of your team’s time - turns the rubric into a written, agreed document. The session opens as an interview with whoever runs this today: how it really works, where it breaks. We connect read-only to Salesforce and Form Assembly, and start building the scoring engine the same week.

Week 2
HMIUS

Score & compare

Based on your past decisions, the agent scores this season’s applications. Your team compares its ranking to their own, side by side. We also test it on past seasons you already decided.

Week 3
HMIUS

Tune, trust & fix

A week of testing. We fix what the team asks for: rubric calls the agent got wrong, and anything in the tool or dashboard that doesn’t work the way you need.

Week 4
HMIUS

Live, with write-back

Scores and rationale flow into Salesforce automatically. Weekly check-in continues. Documentation handed over.

HMI your part · US ours · both badges = together

Every step reversible until you approve write-backThen the same playbook repeats for workflow #2
Workflows 1-4 on the mapWhat this buys back

Your time, back. In numbers.

info@ responses !
“eats hours of staff time a day” - as you put it
25-45hours back, monthly
75-135hours back, quarterly
300-540hours back, yearly
Application review

every application read by hand, 15 trips recruiting. applications/month × 10-20 min × 60-80% saved - we fill the volume together in the rubric session

Document tracking

“heavy, repetitive work,” per couple per trip. standard follow-ups run themselves; volume confirmed in the same session

Grant reporting

37 funders, and an open grant-writer position. the anchor is the role you already budgeted and haven’t filled

Ranges use your numbers where given; assumptions are ours and markedSo quarter two runs on your data, not estimates
Step 2

Applicant emails & document tracking

Workflow 2 · Emails & documentsWhy this link

The paperwork chase.

Today, as you put it

  • Staff chase passports, payments and documents, for every couple and trip
  • The load spikes when several trips onboard at once

Why this spot in the chain

  • Same team as the pilot - the win compounds
  • Runs on the fields and access the pilot already wired

The win

  • The completion dashboard you already asked for
  • It joins the scoring dashboard as a tab
As you put it: no single view of where every couple standsThe dashboard they asked for, in the tool they already have
Workflow 2 · Emails & documentsThe architecture

How onboarding will run.

In place today1 · Couple registersRegistration confirmed, onboarding starts in Salesforce. Already true today.
→
Automation2 · One status tabA dashboard tab shows what each couple finished and what is missing. It joins the scoring dashboard.
→
Automation3 · Reminders send themselvesWhen an item is missing or a deadline nears, the reminder goes out. A stalled couple is raised to staff.
→
Your team4 · Staff handle exceptionsPeople step in only where judgment is needed. The routine chase is over.
→
Automation5 · Back on trackOnce the person steps in, the status updates on the dashboard and the couple returns to automatic tracking.
Same card, same Salesforce, same dashboardOwned by the team that knows this process cold - as you put it
Workflow 2 · Emails & documentsThe build plan

Same playbook, five weeks.

Week 1
HMIUS

Checklist, wire & build

One session with the team: which items - passports, payment plans, the rest - which deadlines, and what a reminder sounds like. The session opens as an interview with whoever runs this today: how it really works, where it breaks. The tab connects read-only, and the build of the tracking tab and reminder engine starts the same week.

Week 2
HMIUS

Compare

Reminder drafts run next to what staff actually send.

Weeks 3-4
HMIUS

Tune & fix

Two weeks of testing. We fix what the team asks for, in the reminders and in the tab.

Week 5
HMIUS

Live

Reminders send, with the approval settings you chose.

The full schedule - early finishes roll straight into the next linkStarts after the scoring link proves itself
Workflow 2 · Emails & documentsWhat it buys back

The payback, measured.

Applicant emails & document tracking !
“Heavy, repetitive work,” per couple, per trip - as you put it.
The scale is already on the table: 15 trips recruiting now, 25 more planned - as you put it. The formula: couples per trip × missing items × minutes per chase; we fill the volumes together in the definition session, and the dashboard then measures the saving for real.
No single view today - as you put itThe pilot’s dashboard measures this from day one
Step 3

info@ responses

Workflow 3 · info@Why this link

The inbox that eats hours.

Today, as you put it

  • “Eats hours of staff time a day”
  • The same questions come up constantly

Why this spot in the chain

  • Daily and visible to the whole org
  • A win everyone feels, every single day

The win

  • Hours back and consistent answers
  • From your own knowledge, not a stranger’s script
Answers depend on who happens to reply - as you put itOne knowledge base, one voice
Workflow 3 · info@The architecture

How info@ will run.

In place today1 · A question arrivesSame inbox, same address. Nothing changes for the couple.
→
Automation2 · The base holds your answersA question-and-answer base, built from replies you already wrote.
→
Agent3 · A draft appearsThe agent drafts a reply from the base, in your tone.
→
Your team4 · A person approvesNothing sends without approval. Routine answers earn lighter review over time.
Consistent answers, from your own knowledge“Hours a day” back - as you put it
Workflow 3 · info@The build plan

Same playbook, faster: four weeks.

Week 1
HMIUS

Interview & mine past emails

A sit-down with whoever answers info@ today: how it really works, what repeats, where it breaks. Then we build the base from the questions you already answered.

Week 1
HMIUS

Read-only drafts

Drafts are written but nothing sends.

Week 1
HMIUS

Compare

Drafts run next to real replies staff wrote.

Weeks 2-3
HMIUS

Tune & fix

Two weeks of testing. We fix tone, accuracy, and anything staff flag.

Week 4
HMIUS

Live, with approval

Every reply still passes a person before it sends.

The full schedule - early finishes roll straight into the next linkStarts after the previous link proves itself
Workflow 3 · info@What it buys back

The payback, measured.

info@ responses
“eats hours of staff time a day” - as you put it
25-45hours back, monthly
75-135hours back, quarterly
300-540hours back, yearly
Named in the meeting: “she spends a few hours a day doing it.” Our marked assumption stays conservative: 2-3 hours a day, and a 50-70% cut in drafting time. The pilot’s baseline replaces this with your measured numbers.
The one link with numbers already on the tableMeasured for real from week one
Step 4

Grant writing and reporting

Workflow 4 · Grant reportingWhy this link

37 funders, one engine.

Today, as you put it

  • 37 funders, each expecting its own format and depth
  • A large, recurring writing load, all by hand

Why this spot in the chain

  • A second team joins after operations wins
  • The visible pilot is what earns this trust

The win

  • The role you budgeted and haven’t filled, largely covered
  • Drafts from your own data and past narratives
Open grant-writer headcount - as you put itThe same data reshaped by hand today, drafted by the engine tomorrow
Workflow 4 · Grant reportingThe architecture

How grant reporting will run.

In place today1 · The data is already yoursProgram outcomes and surveys of roughly 7,000 alumni, in your systems.
→
Agent2 · A draft per funderThe agent drafts each report in that funder’s format and depth.
→
Agent3 · Proposals from your libraryAssembled from your own boilerplate and past narratives.
→
Your team4 · A person approves every reportYour rule, built in: nothing reaches a funder unreviewed.
→
Automation5 · The reports log !A dashboard tab lists every report: funder, name, date, and a status mark - drafted, in review, approved, sent. The whole year at a glance.
37 funders, one drafting engineYour weekly opportunity agent stays as is
Workflow 4 · Grant reportingThe build plan

Same playbook, four weeks.

Week 1
HMIUS

Formats on paper

A session with the development director: funders, formats, depths - and how reporting really runs today, where it breaks.

Week 2
HMIUS

Wire, read-only

The agent reads program and survey data. Nothing is drafted for sending.

Week 3
HMIUS

Real report, tune & fix

A draft runs next to a report you actually sent, side by side - and the same week formats are sharpened funder by funder.

Week 4
HMIUS

Live cadence

Drafts flow on the reporting calendar. A person approves each one.

The full schedule - early finishes roll straight into the next linkStarts after the previous link proves itself
Workflow 4 · Grant reportingWhat it buys back

The payback, measured.

Grant writing and reporting !
The anchor is the grant-writer role you already budgeted and haven’t filled.
37 funders, recurring formats. The formula: reports per period × hours per report × the share drafting takes; give us these in the definition session, and the reports log then measures the saving for real.
A budgeted role as the value anchorEvery report still leaves with human approval
Step 5

Recruitment and marketing

Workflow 5 · Recruitment marketingWhy this link

Marketing, with honest wiring.

Today, as you put it

  • Campaigns built by hand for each city
  • Referral tracking is leaky; warm-lead lists cross-referenced manually

Why this spot in the chain

  • Later in the chain, by design
  • It leans on clean applicant data from steps 1-2 (slides 7 and 12)

The win

  • The warm-lead list builds itself
  • Referral tracking gets a full fix: a form field, the code wire, and AI only for past name lists
Awareness is the core problem - as you put itInfluencer channels first - Avi’s steer
Workflow 5 · Recruitment marketingThe architecture

How recruitment marketing will run.

Automation1 · “Who referred you?” on the form !A simple choice field on the application. The list is local: each couple sees the rabbis and partners of their own city and community. On the custom pages the list arrives pre-filtered. Even word-of-mouth gets counted.
→
No AI2 · Codes stamp the cardA code that was used is written to the couple’s card automatically. A wire, not an AI build.
→
Agent3 · Past names, matchedName lists partners already sent are matched to applicants, spelling variants included. Borderline matches go to a person.
→
Agent4 · The pipeline is scoredEmail engagement, fit and referral signals become a warm-lead call list.
→
Agent5 · Drafts per cityCampaign and outreach versions drafted for each city and audience.
→
Your team6 · Laura & Jen approveStrategy and sign-off stay exactly where they are today.
Influencer channels first - Avi’s steerThe targeting question: answered on the legal slide
Workflow 5 · Recruitment marketingThe build plan

Same playbook, five weeks.

Week 1
HMIUS

Signals, wire & build

One session sets the warm-lead signals, the referral field with the partner list, and the city templates. The session opens as an interview with whoever runs this today: how it really works, where it breaks. The agent scores read-only, the code wire connects, and the build of the lead engine starts the same week. No outreach happens.

Week 2
HMIUS

Compare lists

The agent’s call list next to the one staff built by hand. Past name lists matched as a test.

Weeks 3-4
HMIUS

Tune & fix

Two weeks of testing, on signals, drafts and matching.

Week 5
HMIUS

Live

Lists, drafts and attribution flow. Approval stays with Laura and Jen.

The full schedule - early finishes roll straight into the next linkStarts after the previous link proves itself
Workflow 5 · Recruitment marketingWhat it buys back

The payback, measured.

Recruitment and marketing
Referral tracking is leaky and lists are cross-referenced by hand - as you put it.
Recovered referrals and hours are measured from the first campaign cycle. The attribution gap closes at the source: the form field counts what the codes miss.
Leaky attribution - as you put itMeasured across one registration cycle
Step 6

Personalized applicant pages

Workflow 6 · Personalized pagesWhy this link

A page that knows the couple.

Today, as you put it

  • Every couple lands on the same generic page
  • Drop-off after the homepage is high

Why this spot in the chain

  • Last in the chain, by design
  • It uses the clean data and trust the other links build

The win

  • A page that feels made for them
  • Name, city and trip, already filled in
Paid and earned traffic leaks before it converts - as you put itWorks alongside the login feature Trudy and Laura are building
Workflow 6 · Personalized pagesThe architecture

How personalized pages will run.

In place today1 · A link goes outSame campaigns, same links you send today.
→
Automation2 · The page builds itself !Name, city and trip pulled from Salesforce. It feels made for them.
→
Agent3 · The message adaptsCopy tuned to where the couple is - safety questions answered up front when needed.
→
Your team4 · Brand sign-offDesign and brand stay with your marketing team.
→
Automation5 · The numbers tabA dashboard tab shows how the pages perform: visits, clicks, and application starts, per city and campaign. Same dashboard, one more tab.
Cuts the drop-off after the homepage - as you put itWorks alongside the login feature Trudy and Laura are building
Workflow 6 · Personalized pagesThe build plan

Same playbook, five weeks.

Week 1
HMIUS

Template & brand session

One session with marketing: the template, the tone, the rules - and how pages are built and measured today.

Week 2
HMIUS

Sample pages, offline

Real couples, sample pages. Nothing is public. This is the build of the page engine itself.

Weeks 3-4
HMIUS

Compare, test & tune

Sample pages next to the generic page, on one segment - plus fixes on copy, design and data pulls.

Week 5
HMIUS

Live for one city

One cohort goes first. The measured click-to-application rate decides the rollout.

The full schedule - early finishes roll straight into the next linkStarts after the previous link proves itself
Workflow 6 · Personalized pagesWhat it buys back

The payback, measured.

Personalized applicant pages
Drop-off after the homepage is high - as you put it.
The baseline is the click-to-application rate, before and after, on one city cohort first - measured live in the pages tab of your dashboard.
High drop-off - as you put itThe rollout decision runs on measured numbers
Workflow 5 · Recruitment marketingThe question you asked us to take home

The social-crawling question, answered. !

What’s workable

  • Reaching new couples through the platforms’ official ad tools: interests, life stage, geography
  • Lookalike audiences: Meta finds people similar to couples you already served - their tools, your list, no scraping !

What we’d avoid

  • Scraping individuals’ profiles to build target lists
  • Anything that couldn’t be explained to a couple face-to-face

The better path

  • Avi’s instinct was right: influencer & partner channels first
  • Full written brief follows this meeting
Privacy commitment from our first conversation standsWritten summary to follow
Beyond the map

Our initiatives

Beyond the mapOn our initiative

And four things you didn’t ask for.

When something doesn’t need AI, we say soThat’s the difference between advice and a sales pitch
The pilotBeing straight about it

What this takes, from both sides.

From your team HMI

  • One rubric session with the team that knows the process end to end
  • An owner on your side, a few hours a week during the pilot
  • Review of flagged borderline cases

From us US

  • Build, wire to Salesforce, tune against real applications
  • Weekly check-in during the pilot
  • Documentation your team can run without us

The tradeoffs

  • Narrow on purpose: one workflow, done visibly well
  • The rubric must be agreed before anything is automated
  • Pricing follows in writing right after this meeting

HMI your part · US ours · both badges = together

No open-ended retainer · a defined pilotPricing: in writing, after this meeting
The pilotPricing · how it's built

Priced per part. Fixed. No open meter.

1 · Rubric & setup

  • The working session, the written rubric, agreed edge-case rules
  • Standalone value: you keep it even if we stop here

2 · Build & connect

  • The scoring agent and your dashboard, wired to Salesforce and Form Assembly
  • Read-only until you approve write-back

3 · Run, tune & hand over

  • Five weeks on real applications, weekly check-ins
  • Documentation your team runs without us
The method: every link is a fixed price, sized against the time it buys back, not against hours we spend. One written quote covers the whole chain, within 48 hours of this meeting - you decide once, and we go straight to build. Each link still goes live only after the one before it proves itself.
Rubric changes after calibration = phase 2, priced separatelyQuote in writing, 48 hours
The pilotTo leave this call ready

Three things from you,
and we start.

1

An owner HMI

One person on the HMI side who carries AI. You said it yourselves: whoever you assign matters as much as the tooling.

2

Access HMI

Salesforce and Form Assembly, scoped to the pilot. Read first, write-back when you approve it.

3

One rubric session HMIUS

With the team that knows the process end to end - so the criteria finally live on paper, not in heads.

HMI your part · US ours · both badges = together

Aug 26, 2026

Thank you.

1 / 9