# Epicor Prophet 21 AI Assistant Sizing and Value Calculator

> Source: https://netray.co/tools/epicor-prophet-21-ai-assistant-sizing-and-value-calculator
> Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems.
> Section: Free Tools

## Short answer

Estimate GPUs, index size, hours saved and payback for a private AI assistant used by counter sales and purchasing on Prophet 21.

## Category

epicor

## Erp

Epicor Prophet 21

## Industry

Wholesale Distribution

## Audience

- CIO
- IT manager
- Finance controller
- ERP administrator

## Tool

**Kind:** calculator

**Currency:** $

**Inputs:**

#### Named users

**Key:** users

**Type:** number

**Default:** 160

**Min:** 1

**Step:** 10

**Unit:** users

**Help:** Users licensed to use the assistant.

#### Daily active share

**Key:** active_pct

**Type:** slider

**Default:** 55

**Min:** 5

**Max:** 100

**Step:** 5

**Unit:** %

**Help:** Share of named users who ask at least one question on a given day.

#### Questions per active user per day

**Key:** q_per_user

**Type:** number

**Default:** 14

**Min:** 1

**Max:** 200

**Step:** 1

**Unit:** questions

**Help:** Pilot logs give the real figure; 5 to 15 is common once habit forms.

#### Prompt plus answer tokens per question

**Key:** tokens_per_q

**Type:** number

**Default:** 3500

**Min:** 500

**Max:** 32000

**Step:** 250

**Unit:** tokens

**Help:** Question, rows from the reporting copy and catalogue chunks, plus the answer. Catalogue snippets raise this.

#### Documents to index

**Key:** docs

**Type:** number

**Default:** 90000

**Min:** 0

**Step:** 1000

**Unit:** documents

**Help:** Price sheets, supplier catalogues, spec sheets, contracts and order PDFs.

#### Average pages per document

**Key:** pages_per_doc

**Type:** number

**Default:** 6

**Min:** 1

**Max:** 500

**Step:** 1

**Unit:** pages

**Help:** Average across the document set.

#### Minutes saved per question

**Key:** minutes_saved

**Type:** number

**Default:** 3

**Min:** 0

**Max:** 120

**Step:** 0.5

**Unit:** minutes

**Help:** Counter lookups save minutes, not hours; use a smaller figure than manufacturing planning questions.

#### Loaded hourly cost

**Key:** hourly_cost

**Type:** number

**Default:** 38

**Min:** 0

**Step:** 1

**Unit:** per hour

**Help:** Salary plus benefits and overhead for the people asking.

#### GPU monthly cost

**Key:** gpu_monthly

**Type:** number

**Default:** 1200

**Min:** 0

**Step:** 100

**Unit:** per GPU per month

**Help:** Hardware amortised over 36 months plus power, cooling, support and rack space.

#### Throughput per GPU

**Key:** tps_per_gpu

**Type:** number

**Default:** 1200

**Min:** 100

**Max:** 10000

**Step:** 100

**Unit:** tokens/s

**Help:** Aggregate tokens per second under batching for your model and quantisation. Measure it; this is the most sensitive assumption.

#### One-time implementation cost

**Key:** impl_cost

**Type:** number

**Default:** 50000

**Min:** 0

**Step:** 5000

**Unit:** one-time

**Help:** Integration, document pipeline, security review and pilot.

**Outputs:**

#### Questions per day

**Key:** q_day

**Expr:** users * active_pct / 100 * q_per_user

**Format:** integer

**Description:** Named users x active share x questions per active user.

#### Tokens per day

**Key:** tokens_day

**Expr:** q_day * tokens_per_q

**Format:** number

**Description:** Questions per day x tokens per question.

#### Peak concurrent requests

**Key:** peak_conc

**Expr:** max(1, ceil(q_day * 0.25 / 3600 * 15))

**Format:** integer

**Description:** Assumes 25 percent of daily questions land in the busiest hour and each request holds a slot for about 15 seconds.

#### GPUs needed (before redundancy)

**Key:** gpus

**Expr:** max(1, ceil(q_day * 0.25 / 3600 * tokens_per_q / tps_per_gpu))

**Format:** integer

**Primary:** true

**Description:** Peak-hour tokens per second divided by throughput per GPU, rounded up. Add a spare GPU if you need failover.

#### Vector index size

**Key:** index_gb

**Expr:** docs * pages_per_doc * 2 * 1024 * 4 * 1.5 / 1000000000

**Format:** decimal

**Description:** Gigabytes, assuming 2 chunks per page, 1024-dimension float32 embeddings and 50 percent index overhead. Original files are extra.

#### Hours saved per month

**Key:** hours_saved

**Expr:** q_day * minutes_saved / 60 * 21

**Format:** hours

**Description:** Questions per day x minutes saved, over 21 working days.

#### Monthly value

**Key:** value_month

**Expr:** hours_saved * hourly_cost

**Format:** currency

**Primary:** true

**Description:** Hours saved x loaded hourly cost. Treat as an upper bound until a pilot confirms the time saved.

#### Simple payback

**Key:** payback

**Expr:** if(gt(value_month - gpus * gpu_monthly, 0), impl_cost / (value_month - gpus * gpu_monthly), 0)

**Format:** months

**Description:** One-time cost divided by monthly value minus GPU cost. Shows 0 when value does not exceed GPU cost.

**Chart:**

#### GPUs needed vs tokens per question

**Type:** line

**X:** tokens_per_q

**From:** 1000

**To:** 16000

**Steps:** 31

**Series:**

- gpus

**Disclaimer:** Planning estimate only. Tokens per second, time saved per question and peak share vary widely; replace the defaults with measurements from your own pilot before committing to hardware.

## Content

**Intro:** This calculator sizes a private AI assistant on Prophet 21 for counter sales and purchasing. It estimates GPUs for the busiest hour, vector index size, hours saved and payback.

**Sections:**

#### How to use it

**Body:** Enter named users and the share active daily; distribution counters ask more questions than planners, so use a higher questions-per-user figure from a pilot if you can. Tokens per question grow when catalogue chunks are retrieved. Use minutes saved per lookup that matches counter work, which is usually a few minutes.

**Bullets:**

- Counter users can ask 10 to 25 questions a day
- Short answers keep tokens low
- Treat branch opening hours as the peak
- Compare to P21 screens that already answer the question

#### The method

**Body:** Daily questions = users x active share x questions each. Busiest hour carries 25 percent of the day. Tokens per second at peak = questions x 0.25 / 3600 x tokens per question. GPUs = that rate divided by throughput, rounded up. Index = documents x pages x 2 chunks x 1024 x 4 bytes x 1.5. Value = daily questions x minutes saved / 60 x 21 x hourly cost.

#### Reading the sweep

**Body:** The line shows how GPU need rises with context length. Retrieval-heavy answers over catalogues use far more tokens than simple availability lookups, so keep the context tight. If the line crosses to a second GPU early, consider a smaller model or answer-length limits before buying hardware.

#### Common mistakes

**Body:** Common errors include overestimating minutes saved per lookup, ignoring peak-hour concentration at branch open, and counting the value of questions the P21 screen already answers quickly. Use the assistant where it spans items, customers and vendors.

**Cta:** If you want these numbers tested against your branch usage, Netray can help with a pilot. The calculator works without any contact.

**Faq:**

#### Why are counter minutes saved so low?

**Answer:** Counter questions are quick lookups, so the saving per question is a few minutes at most. The value comes from volume. Use a conservative figure and let high question counts drive the total.

#### Does branch count change the result?

**Answer:** Not directly; users and question volume do. But branches that open at the same time create a sharper peak, so use the busiest hour from your own logs where you can.

#### Is the index size large for distributors?

**Answer:** It can be, because catalogues and spec sheets are page-heavy. The estimate covers vectors only. Plan separate storage for the files, the extracted text and the SQL reporting copy.

#### Can I use this for budget approval?

**Answer:** Yes as a first estimate, with a note that inputs are assumptions. Replace defaults with pilot measurements for throughput, question volume and minutes saved before presenting the final case.

## Search terms

Prophet 21 AI sizing, P21 AI ROI calculator, distribution AI assistant cost, Prophet 21 GPU sizing, wholesale distribution AI payback, on-prem AI sizing calculator

---
Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json