Graity — Implementation Guide for Business & Enterprise

The Problem

88% "use AI". ROI is rare.

McKinsey, Gartner and BCG all agree: companies buy AI tools at scale — and see no financial return. The reason is always the same: knowledge is scattered, unstructured and inaccessible to any AI system.

70% of AI transformation complexity is people and process, not technology (BCG). You can't run intelligent agents on chaos. Before Orbit can act, Core must be built.

74%

stuck at pilot stage (BCG)

1/50

reach real AI transformation (Gartner)

60%

see no value from AI spend (BCG)

57%

data not enterprise-grade (Gartner)

Graity solves this at the root level. Core is the knowledge infrastructure — the AI-ready memory layer every agent and workflow needs. Orbit is the execution layer built on top of it. Without Core, Orbit has nothing to think with.

The root cause

Knowledge lives in files nobody reads
Experts leave and take know-how with them
AI runs on internet data, not company reality
No single source of truth anywhere

Graity Core fixes

Collects every format automatically
AI-processes and vectorises all content
Stays current with live sync & triggers
Becomes the company's searchable memory

Graity Orbit enables

Custom AI agents for any workflow
700+ service integrations in one place
Trigger-based process automation
AI that acts on real company data

Maturity Levels

Where are you now?

Gartner defines 5 AI maturity levels. Identify yours before starting. Each phase of the Graity deployment guide is mapped to a maturity transition.

L1
Ad Hoc
Chaos — no AI strategy
Individual employees use ChatGPT privately. No governance, no shared knowledge, no visibility. Most companies are here.
Start: Graity Core deployment
L2
Basic
First pilots — tools but no system
A few AI tools in use. Pilots with no scale. No unified knowledge base. Results are person-dependent, not systemic. 37% of companies can't even pick which use case to start with.
Focus: Core is live, first agents in Orbit
L3
Standard
Process & governance in place
Graity Core is the single source of truth. Orbit automates 3–5 core workflows. KPIs are tracked. Governance is active. First measurable EBIT impact.
Focus: Reshape 3–5 key processes end-to-end
L4
Scale
Cross-functional, self-improving
Knowledge base grows autonomously. Agents write back to Core. Multiple departments use Orbit. External interfaces (client bots) are live. System learns continuously.
Focus: Orbit agents across departments + external interfaces
L5
Adaptive
AI in the DNA of the organisation
Every employee is a manager of AI agents. Core and Orbit are operational infrastructure, not tools. New products and services built on company intelligence. This is where 2–6× TSR advantage is created.
Continuous: Invent — new business models

Deployment Roadmap

Six phases. One direction.

Structured after McKinsey's Rewired framework, Gartner's 7 workstreams and BCG's Deploy → Reshape → Invent model — mapped precisely to Graity. Click each phase to expand.

00

Audit & Prioritise

what knowledge exists · what processes to automate first

Week 1–2

▶

$

Map all knowledge sources. List every place where company knowledge lives: shared drives, email threads, recorded meetings, CRM notes, Confluence/Notion, WhatsApp chats, offline expertise in people's heads. This is what Core will ingest.

$

Prioritise what to feed Core first. Use the matrix: high reuse × high search frequency = ingest first. Start with: SOPs, client briefs, meeting transcripts, product documentation, onboarding materials.

$

Map processes for automation. Rate each routine process on two axes: frequency × cost of error. High on both = automate first with Orbit. Examples: meeting summaries, report generation, client Q&A, competitive monitoring.

$

Identify connectors needed. Which tools does your team already use? CRM, email, Slack, project management, accounting? These become Orbit connectors. List them — they're ready to integrate.

>

Define success metrics upfront. Before deploying anything, agree: What does success look like in 90 days? KPIs might include: hours saved per week, response time to client queries, % of onboarding handled by AI, search queries answered without human input. Without this, you cannot measure ROI (McKinsey: KPI tracking is the #1 EBIT driver).

Output Knowledge source inventory · Process priority matrix (2×2: routine × error cost) · Connector list · 90-day KPI baseline

01

Core Deployment

knowledge infrastructure · data ingestion · RAG base

Week 2–4

▶

$

Connect primary data sources to Core. Start with the highest-value sources from Phase 0. Upload PDFs, connect Google Drive/Docs, link external documentation URLs. Core ingests all formats — each passes through the 5-layer AI pipeline: OCR → parse → enrich → deduplicate → vectorise.

$

Set up live sync for dynamic sources. Google Workspace documents get auto-update triggers. Core will re-index whenever a document changes. If access is ever lost, the last version is preserved. Knowledge becomes source-independent.

$

Ingest audio and video expertise. Upload recordings of key meetings, workshops, client calls. Core uses Whisper transcription → standard text pipeline. Expertise locked in audio is now searchable and queryable.

$

Configure access permissions. Set visibility scopes: company-wide, department, team, individual. Use HR role definitions. One Core, multiple controlled "lenses" of reality. Critical for enterprise governance (Gartner TRiSM layer).

$

Validate the knowledge base. Run test searches across the ingested content. Check that the AI retrieves correct, contextually relevant results. Identify gaps — these become the input for Phase 2's Orbit gap-detection agent.

Output after Phase 1 Graity Core is live. All priority knowledge is ingested, vectorised and searchable. Access scopes configured. The company now has its first unified, AI-ready knowledge base. From this point, every AI agent has real data to think with.

02

Orbit Activation — First Agents

AI chat · first connectors · quick wins

Week 3–6

▶

$

Deploy the AI Chat interface for the whole team. This is BCG's "Deploy" phase — roll out to the entire organisation. Every employee gets one window to query Core, choose a model and scope their search. Goal: +10–15% productivity, excitement and evidence that AI delivers. No complex setup required.

$

Connect your first 3 external services via Orbit. Start with what your team already uses most: email (Gmail/Outlook), CRM, Slack or Telegram. These become live data inputs for Core and execution targets for agents.

$

Build your first 2 agents from the process priority list. Choose the top 2 from Phase 0: highest frequency + tolerable error cost. Examples: Meeting Intelligence Agent (transcription → task extraction → CRM update) or Brief-to-Content Agent (brief → Core context → draft). These are your first "Reshape" experiments (BCG framework).

$

Connect Core's gap-detection. Configure Orbit to identify questions asked via Chat that returned no answer from Core. These become a queue of knowledge gaps for moderated addition. The system begins self-identifying what it doesn't know.

>

Track KPIs from day one. Measure hours saved, queries answered autonomously, outputs generated. Show the numbers to the team. McKinsey: this is the most critical adoption driver — visible wins create trust, trust drives usage, usage drives value.

Output after Phase 2 Orbit is live with 2 working agents. Team is using AI Chat daily. First connectors active. KPIs tracked. Early productivity gains visible and measured. Quick wins fund the next phase.

03

Workflow Redesign

reshape · end-to-end automation · EBIT impact

Month 2–4

▶

$

This is McKinsey's #1 EBIT driver: workflow redesign. The critical distinction: "AI on top" means a manager still fills 15 CRM fields manually but AI suggests the text. Workflow redesign means the manager speaks a meeting summary → AI fills the CRM, creates tasks and sends the follow-up. The process is rebuilt, not patched. Target: +30–50% efficiency in redesigned functions (BCG).

$

Select 2–3 core business processes for full end-to-end Orbit automation. Not back-office. Core functions: client communication flow, content/campaign production, reporting pipeline, research and competitive intelligence, onboarding. BCG: 62% of AI value comes from core functions, not support.

$

Build multi-step Orbit flows for each process. Example — Performance Report Flow: weekly trigger → pull data from CRM + analytics APIs → compare against targets stored in Core → generate formatted report → send to Slack and save output to Core. Every step is connected, automated, and self-logging.

$

Expand Core continuously. Every agent output that contains new knowledge is reviewed and added to Core. Meeting summaries, research findings, client feedback, competitive intelligence — all flow back. The knowledge base grows with every workflow cycle.

>

Governance: activate moderation layer. Before any AI-generated content enters Core, set moderation review (human-in-the-loop). Configure TRiSM levels: who can see what, which agents can write where, audit log for all actions. Gartner: 80%+ of AI incidents are internal, not from external attacks.

Output after Phase 3 First measurable EBIT attribution. 2–3 workflows fully automated. Core growing autonomously. Governance active. System begins delivering compound value — each cycle makes the next one smarter.

04

Scale Across the Organisation

departments · external interfaces · product layer

Month 4–9

▶

$

Roll out Orbit agents to additional departments. Marketing → content and SEO agents. Sales → CRM update, lead research, pitch prep agents. Operations → meeting intelligence, reporting agents. HR → onboarding assistant, policy Q&A bot. Each department gets scoped access to Core and their own Orbit workflows.

$

Deploy external interfaces. Activate client-facing bots: Telegram or Slack bots that query the client's scoped RAG segment. Choose mode: "Librarian" (surfaces references and links) or "Consultant" (full AI dialogue). Company knowledge becomes a client-deliverable product.

$

Assetise successful workflows. McKinsey: convert proven agents and flows into reusable templates. Target 60–90% reuse, 10–40% customisation per new deployment. One well-built Meeting Intelligence Flow should serve Sales, Operations and HR with minimal modification.

$

Expand connector coverage. Move beyond initial 3 connectors. Add MCP tools for deeper agent action: file management, calendar, project boards, accounting systems. Each new connector multiplies what Orbit agents can do.

>

Measure ROE (Return on Employee) alongside ROI. As Gartner defines it: how much value does one person create with AI augmentation? This is the metric that reflects true productivity transformation, not just cost savings.

Output after Phase 4 Graity Orbit deployed across all key departments. External interfaces live. Client knowledge products operational. Core is self-growing. The system functions as a continuous intelligence layer — not a set of tools.

05

Continuous — AI in the DNA

invent · new business models · self-learning

Month 9+

▶

$

This is not a project with a finish line. 45% of mature companies have kept AI in production for 3+ years (Gartner). Plan budget and capacity on a 3–5 year horizon. Monthly review of KPIs. Quarterly strategy revision — the AI market changes faster than any annual plan.

$

Launch "Invent" initiatives (BCG framework). Use accumulated Core intelligence to build new products and revenue streams. Examples: a knowledge-as-a-service product for clients, AI-powered research offering, automated content packages based on your expertise corpus.

$

Core becomes self-auditing. Orbit agents actively scan for knowledge gaps — questions that returned no answer — and surface them with suggested sources. A human moderator reviews and approves additions. The knowledge base identifies its own blind spots.

$

Continuous governance and compliance. Audit log review. Access scope review as team structure changes. Monitor token usage by department and agent (CPT metric). Governance is not a setup task — it's an ongoing operational function.

>

Target state: Gartner L5 Adaptive. Every employee manages AI agents. Core and Orbit are operational infrastructure, not tools anyone "implements." New colleagues are onboarded by AI. Client intelligence is generated continuously. The organisation thinks and learns as a single organism.

Target state Company knowledge → competitive moat. Core is the institutional memory that grows with every project, call and document. Orbit is the execution layer that acts on it. Together: the operating system for the organisation's intelligence.

KPIs & Metrics

What to track. When.

McKinsey: KPI tracking is the single highest-impact practice for EBIT improvement. Without measurement, you cannot scale. Without scale, there is no transformation.

Metric	Target	What it measures
Hours saved / week	5–15h per employee	BCG: AI returns 26–36% of employee time. Baseline against Phase 0 audit.
Queries answered autonomously	>70% within 60 days	Measures Core coverage quality. Low score = more ingestion needed.
Core coverage rate	Growing weekly	% of team questions Core can answer without human escalation.
Workflow cycle time	−60–80% vs baseline	McKinsey workflow redesign benchmark. Compare pre- and post-Orbit automation.
Agent output → Core additions	Positive weekly trend	Measures self-learning loop. Each cycle should add knowledge.
EBIT attribution	+20%+ in target functions	McKinsey: the ultimate success benchmark. Ties AI spend to financial outcome.
ROE (Return on Employee)	Track quarterly	Gartner metric: how much value per person with AI augmentation. Shows productivity transformation, not just cost savings.
CPT (Cost Per Token)	Declining trend	AI FinOps foundation. Measures AI infrastructure efficiency. Drives adoption of local models and routing optimisation.

Common Failures

What kills AI transformation.

Each item below is a real cause of failure — documented across McKinsey, Gartner and BCG research. Read before you start.

// deploy Orbit without Core

Agents built on empty or unstructured data hallucinate and deliver wrong answers. Core is not optional — it is the foundation. Every Orbit action is only as good as the knowledge behind it.

// AI on top, not redesign

Adding AI suggestions to an unchanged process gives +5–10% gains. Redesigning the process around AI gives +30–50%. McKinsey: workflow redesign is the #1 EBIT driver, not tool adoption.

// 15 pilots, no depth

When AI transformation stalls, 80% of cases recover by dropping scattered experiments and focusing on 2–3 real business processes. Depth beats breadth at every stage.

// governance "later"

TRiSM cannot be retrofitted. Set access scopes, audit logging and moderation in Phase 1. 80%+ of AI incidents are internal, not from external attacks — your own team is the risk.

// no KPIs before deploying

Without baseline metrics from Phase 0, you cannot prove or improve ROI. Every deployment needs a "before" and "after" measurement. If you cannot measure it, you cannot scale it.

// one department only

If Core is used by one team, the knowledge base is narrow. The value of Graity compounds as coverage grows — each department adds knowledge that every other department can use.

// static knowledge base

A knowledge base that isn't updated becomes dangerous faster than it becomes useless. Enable live sync, update triggers and the gap-detection loop from day one. Core must always reflect current reality.

// waiting for "readiness"

BCG: start small, scale fast. Deploy funds Reshape. Every organisation has enough knowledge to start Core in week one. Perfect data architecture is a myth — start with what you have, improve continuously.

Deployment Track

SMB or Enterprise?

Same products, different starting points. The fundamental difference: companies start with a tool and discover the process. Enterprises start with the org design and then deploy the tool.

SMB / AGENCY ENTERPRISE / CORPORATION Starting point tool → find the process org design → deploy tool Phase 0 focus map own knowledge corpus assess 7 Gartner pillars first Core scope at launch 1 team, 50–200 documents 1–2 departments, controlled pilot First agents 2 agents in 3 weeks 2 agents in 6–8 weeks (sign-off) First ROI 6–8 weeks 3–6 months Budget — pilot phase $2k–15k / month $50k–200k / month Governance basic access scopes TRiSM 4 layers + AI officer People required train existing team AI lead + change management Scale timeline 3–6 months to L3 12–24 months to L3 Key risk not starting starting without C-suite alignment Core coverage at L3 1 team fully indexed 3–5 departments indexed

Enterprise note

For corporate deployments: the #1 barrier is not technology — it is leadership alignment. McKinsey: CEO governance oversight is the single highest-correlation factor with AI EBIT impact. Before Phase 0, secure C-suite commitment with a specific KPI contract: target metric, timeline and accountability. Without this, the project will stall at pilot stage — as 74% do (BCG).

What Graity Provides

Infrastructure, not advice.

Every element of this guide maps to a specific Graity capability. Nothing in this roadmap requires building or buying additional technology.

Core — every phase

Universal data ingestion (Phase 0–1)
5-layer AI processing pipeline
Auto-deduplication & live sync
RAG-ready vector storage (Qdrant)
Granular access control (TRiSM)
Self-auditing gap detection (Phase 3+)

Orbit — phases 2–5

AI Chat with model selection
700+ connector integrations
Custom agent builder (no code)
Trigger-based automated flows
External interfaces (client bots)
Full audit log + token tracking (CPT)

Governance — built-in

Role-based access from HR systems
Moderation layer before Core ingestion
Usage analytics per user / agent
Data stays in your infrastructure
Audit trail on all agent actions
Segmented client RAG environments

From knowledge chaosto institutional intelligence.