BH Technology Book a consultation

ROI for AI

AI gives you hours back. Where you put them decides what you earn.

Hypothetical, illustrative example. Assumptions are shown in the revenue section.

  1. Hours back100 hfreed each month
  2. Useful activity40 hput into new proposals
  3. Sales result5 clients wonfrom 20 extra proposals
  4. Possible profit$15,000gross profit only if all 5 are served, before AI costs

Step 1 · Capacity

One person moves faster. A department frees more hours.

Example values only. The bars show one typical setup for standard sales proposals. Real results depend on the business. Tap a setup label on either card to see the full stack.

Same bar scale on both cards · hours 0 to 40 h a week · more proposals 0 to 50%

One person

Example · 20 approved proposals a week

  • Hours saved per week10 h
  • More proposals with the same staff time+50%

20 → 30 approved proposals a week

Setup

Percentages compare approved proposals a week at the same approval standard. Today's volume takes fewer hours. Put those hours back into proposals and output can rise. It is the same time used one way or the other, so count it once.

Example values, with the full worked example in each setup overview. Ongoing support for the whole stack is a separate service line on the Budget page.

Capacity is room to grow. Payroll stays the same when an hour frees up. Put each freed hour on qualified follow-ups, proposals or client delivery. Count deals or margin only after that work converts and is delivered. Extra work with no financial result is still capacity.

Measure it · three checks on one page

  1. What changed? How to measure hours freed
  2. Was the time used? Hours assigned and work completed
  3. What limited the upside? Demand, delivery, quality, cost

One person · typical setup example

Overview

Example values for one person who prepares standard sales proposals on their own accounts. A person approves every proposal before it goes out.

Worked example

  1. Today20 approved proposals × 90 min = 30 h
  2. With the setup20 proposals × 60 min = 20 h
  3. Same volume10 h saved a week
  4. Time reinvested30 h at 60 min = 30 proposals · 20 → 30 = +50%

The 60 min is an assumed all-in average at steady state with review, corrections and upkeep included. More output needs enough qualified work and the same approval standard. These are two uses of the same time and a business can also split it.

How one proposal moves

  1. Deal in HubSpotProposal stage reached
  2. Approved sourceTemplate, pricing and examples in Drive
  3. Agent draftsGrok Bot drafts and organizes recurring steps
  4. Person checksClaude compares the draft with approved examples
  5. Person approvesOwner signs off and sends
  6. RecordedNote and file link saved in HubSpot

Who does what each day

  • Grok BotReads the approved queue or an exported deal and drafts proposals and recurring steps
  • ClaudeWrites longer pieces and reviews drafts against approved examples the owner provides
  • CodexBuilds and tests small scripts and reports in VS Code or a repo
  • HubSpotHolds deals, tasks and customer status
  • DriveHolds approved docs and working files

One deal, start to finish

  1. A deal in HubSpot reaches the proposal stage. The owner exports the deal details or adds them to the approved queue.
  2. Grok Bot reads that entry plus the approved template and pricing from Drive, then drafts the proposal and a checklist of recurring steps.
  3. The owner hands Claude the draft and approved examples as files. Claude flags gaps and differences. The owner checks facts, scope, pricing and any exceptions.
  4. The owner approves and sends the proposal. The note, file link and next task go back into HubSpot.
  5. When the routine needs a fix, the owner uses Codex in VS Code to update the script and tests it before the next run.

The owner stays in charge. They hold every account, review each draft and approve what goes out. Tools hand work over by file, link or repo commit.

Entry $30 Grok Premium + $20 ChatGPT Plus + $20 Claude Pro = $70 a month. Heavier tiers $300. New Workspace adds $16.80. CRM and extra usage are separate.

USD before tax. Sources xAI, ChatGPT, Claude.

Go deeper

Department of 10 · typical setup example

Overview

Example values for a department of 10 after rollout, preparing standard sales proposals on company accounts. Three people pilot it first. An assigned owner approves every proposal.

Worked example

  1. Today100 approved proposals × 120 min = 200 staff-hours
  2. With the setup100 proposals × 96 min = 160 h
  3. Same volume40 h saved a week across the team
  4. Time reinvested200 h at 96 min = 125 proposals · 100 → 125 = +25%

The 96 min is an assumed all-in average at steady state with review, corrections and upkeep included. More output needs enough qualified work and the same approval standard. These are two uses of the same time and a business can also split it.

How one proposal moves

  1. Deal in HubSpotStage and owner set
  2. Shared approved contextTemplates, pricing and examples from Drive
  3. Postgres queueProposal tracked by ID with an assigned owner
  4. Agent preparesGrok Bot drafts the proposal and evidence pack
  5. ChecksClaude review against approved examples
  6. Person approvesAssigned owner or manager signs off
  7. RecordedHubSpot status and run log updated

The stack

Records
HubSpot is the source of record, Supabase Postgres tracks the queue, Google Workspace holds approved docs
Agents
Grok Bot Teams seats on company accounts
Builders
Codex through ChatGPT Business, in VS Code and a repo
Reviewers
Claude Team for results and playbooks

One deal, start to finish

  1. A deal in HubSpot reaches the proposal stage and a work item enters the queue with an assigned owner.
  2. The queue service gathers the deal record, the shared approved template, pricing and examples.
  3. Grok Bot prepares a proposal draft and an evidence pack that lists the sources used.
  4. The owner hands the draft and evidence to Claude for a check against approved examples, then checks facts, scope, pricing and exceptions. A manager handles anything unusual.
  5. The owner approves and sends. The approval is logged and the status goes back to HubSpot.

People and checks

  • Team trainingThree pilot operators learn the routine first. All ten join only after the pilot shows evidence.
  • Quality checksEvery draft gets a human check before it goes out. Managers approve templates and exceptions.

The service delivers the working setup. BH Technology maps the workflow, cleans the source data, sets permissions per identity, builds the queue service, tests it in the pilot and trains the team. Each part is scoped and quoted separately.

Example monthly subtotal for licenses and the database is $545 (10 Grok Bot seats, 3 Codex through ChatGPT Business, 3 Claude Team, Supabase Pro). With new Google Workspace it is $713. Usage, CRM, hosting and the setup service are separate costs.

USD before tax. Sources xAI Teams, ChatGPT, Supabase.

Go deeper

Step 2 · Leaders first

Equip the department head first

A lead with AI sees the team's work, coaches sooner and spreads what works. Pilot with the head of the department your assessment selects.

  1. 1. See activity

    Proposals sent this week

    • Rep A6
    • Rep B5
    • Rep C 4 stalled1

  2. 2. Coach and set priority

    Rep C: follow up on the 4 stalled proposals today.

    Whole team: use Rep A's follow-up template.

  3. 3. Same team measure after rollout

    Team proposals sent per week

    • Before12
    • After18

Illustrative values. Team rollout goes smoother once the lead already works this way.

Step 3 · Revenue work

Point freed hours at work that earns

Sales

AI-assisted action
Drafts proposals and follow-ups
Activity measure
Proposals sent, follow-ups done
Business result
New clients and revenue booked

Delivery and service

AI-assisted action
Speeds onboarding and catches rework early
Activity measure
Days to onboard, rework rate
Business result
More work delivered and clients retained

Finance

AI-assisted action
Shows margin by service and staff hours per job
Activity measure
Margin per service, hours per job
Business result
Clearer pricing and staffing decisions

Tracked separately: payment speed, from invoice sent to cash received.

Worked example: Sales

  • 5 reps, 50 proposals a month
  • 4 h → 2 h per proposal
  • 100 h freed
  • 2 h per new proposal
  • 25% close rate
  • $10,000 per client
  • 30% gross margin
  1. 100 hfreed
  2. 40 hput into proposals
  3. 20extra proposals
  4. 5new clients
  5. $50,000booked
  6. $15,000gross profit if all 5 are served
$15,000 gross profitAI setup and running costsNet return

Hypothetical example. Confirm delivery capacity and quality before counting the profit.

Step 4 · Bottlenecks

Fix one area and the bottleneck moves

Before: Delivery unchanged

  • Sales wins5 clients
  • Delivery can start2
  • Waiting3

2 started, 3 waiting

After: Delivery equipped

  • Sales wins5 clients
  • Delivery can start5
  • Waiting0

5 started, 0 waiting

Before sending more work downstream: check delivery capacity, quality and real demand.

Billing, tracked separately: late invoices slow cash. Watch days to payment on their own.

Illustrative values.

Step 5 · Our assessment

How BH Technology assesses

  1. Map where money is madeSales, delivery, service and finance
  2. Read the dataTeam activity plus financials
  3. Pick the first priorityWhat holds back revenue or margin
  4. Equip the lead, then the teamStart with that department's head
  5. Measure the resultOutput, quality, actual cost and margin
  6. ReassessFind the next constraint

Back to step 1 with the next constraint

Find your first priority

We start with the activity that can change a business result.

Book a consultation