Report Overview

Overview

Company has engaged PressW to bring AI and automation into Company.

PressW works in 3 stages - Discovery, Implementation, and Maintenance.

This report is the culmination of this first phase, the Discovery Phase. Within this report, PressW will break down Company’s business and examine each department through the lens of AI. Contained within will be an analysis of each department, a menu of potential AI/automation solutions, a guide on off-the-shelf solutions we believe will apply, and finally our recommendation for a path forward.

Company’s Departments and Stakeholders

Engineering

Key Stakeholders (for this purpose we use one-letter initials instead of names)

• H
• S
• B
• Y - QA
• M - UI/UX
• S - Architecture

Product

Key Stakeholder: M

Sales

Key Stakeholders

• M
• M

Customer Success

Key Stakeholders: C and M

Marketing

Key Stakeholder: S

Menu

Below you will find our menu of project offerings, along with problem statements, descriptions, estimated timelines, and pricing.

📚🤖 Knowledge Base and Q&A Bot

The Identified Issue

Company’s knowledge currently lives in various silos spread across the company.

Knowledge Silo Examples

This knowledge segmentation makes it difficult to find, reference, and build upon Company’s knowledge base. A lot of time and resources are spent trying to locate these pieces of knowledge, and the silos also make it difficult to effectively build out a (robust) knowledge base over time (for example: knowledge lost in dozens of Slack threads and conversations), no transparency and cross-pollination.

The lack of a centralized knowledge repository also makes it difficult to build AI tools which rely on Company’s knowledge. Currently, this lack of centralization blocks many AI applications such as question answer systems, chatbots, and content writers.

Proposed Solution

<aside> ⭐ Aggregate knowledge into a centralized location and build custom LLM tools on top.

</aside>

Project Specifications

• A vector database that stores:
  • Technical Documentation
  • Slack messages and threads
  • Confluence Docs
  • Any other document based knowledge
• A LangChain pipeline and backend service
  • Supports creation of LLM agents that are able to effectively pull from knowledge store
  • Pulls relevant data from vector database and uses effective prompts to guide output
• Slack bot
  • Attempts to preemptively answer questions, or reference documentation, in channels

Reasoning

We picked Slack as the interface for the MVP for this chatbot because:

• Can be utilized internally to test effectively
• Can immediately provide business value by helping answer customer questions saving time and resources
• Most businesses Company works with have their own Slack channel
  • Can be deployed individually

Initially this bot would only work by pulling knowledge from the specific customer Slack channel it is apart of and the knowledge base itself, but can later be extended to support more diverse input knowledge sources.

Timeline and Project Costs

A majority of this cost is in building out your knowledge base, which effectively pulls and aggregates documents from the variety of sources Company uses. This is important to be done correctly, but will become the cornerstone for any new language-based AI/ML solutions Company implements for the rest of time.

Future Additions

Onboarding and Customer Support Bot

• Helps guide customers through onboarding
• Helps answer product questions or direct customers to documentation
  • Can even be extended to allow for actions to be taken through the interface

💸📈 Spend Optimizer

The Identified Issue

Right now Company’s product saves customers money by reducing the number of dev-ops/infrastructure staff needed to run an effective Kubernetes stack. There isn’t anything that is baked into the product that actually helps optimize the customer’s stack itself and save money on cloud spend.

This is an opportunity for Company to massively increase its value proposition to customers. Right now it is extremely difficult to manage cloud spend across Kubernetes clusters, particularly if you are multi or hybrid cloud user. Company is uniquely positioned due to the data being collected to offer spend optimization to its customers and upsell this feature to its customers.

In addition this feature can be used to optimize Company’s internal infrastructure as well, resulting in spend save.

Proposed Solution

<aside> ⭐ A feature within the Company product that analyzes a customer’s Kubernetes setup and automatically suggest cost-saving configuration updates. Continuously monitors spend and optimizes over time.

</aside>

This project would rely on the tremendous amount of data that Company has collected across their customer’s Kubernetes deployments. This would be a large scale feature development that would rely on multiple streams of input data, as well as close communication with Company’s product to execute the proposed changes

For this project, we recommend starting with an MVP build that works on Company’s data alone, before scaling up to handle all customer’s data.

Project Specification

Spend Forecasting Model

• We need the ability to forecast spend, and it must generalize to multitudes of cluster configurations, environments, and cloud providers