Data strategy and engineering for agentic workflows

by Ray Grieselhuber

Embed

Start on current slide

Slide 1

Slide 1 text

HEADSHOT HERE Ray Grieselhuber DemandSphere Data strategy & engineering for agentic workflows @raygrieselhuber

Slide 2

Slide 2 text

No content

Slide 3

Slide 3 text

No content

Slide 4

Slide 4 text

Let’s talk about data strategy & engineering in an agentic world

Slide 5

Slide 5 text

Many of the same engineering principles apply in both product and data engineering

Slide 6

Slide 6 text

Today we’re building a map, don’t worry too much about knowing it all

Slide 7

Slide 7 text

Takeaway: A framework to help you build more reliable products

Slide 8

Slide 8 text

Our goal is to promote more ﬂuency in building repeatable systems with AI

Slide 9

Slide 9 text

Help decision-makers understand what it takes to support AI-engineered products

Slide 10

Slide 10 text

Why am I talking about this at an SEO conference?

Slide 11

Slide 11 text

Agentic engineering can solve problems that have plagued SEO teams for decades

Slide 12

Slide 12 text

Our clients are doing things in months that would have formerly taken years

Slide 13

Slide 13 text

Examples

Slide 14

Slide 14 text

No content

Slide 15

Slide 15 text

Good search marketing is good product management and vice versa

Slide 16

Slide 16 text

Performance in search is a leading indicator of the eﬀectiveness of your corporate strategy

Slide 17

Slide 17 text

Agentic engineering enables the ability to improve both

Slide 18

Slide 18 text

Software as a Service Or Service as a Software?

Slide 19

Slide 19 text

Experience at Functionize taught me about the new SaaS

Slide 20

Slide 20 text

The dirty little secret in SaaS

Slide 21

Slide 21 text

Silicon Valley is ﬁnally starting to ﬁgure this out

Slide 22

Slide 22 text

This is why SaaS isn’t “dead”

Slide 23

Slide 23 text

Side note: “X is dead” is a phrase taught to startup founders to aid in fundraising

Slide 24

Slide 24 text

Vibe coding vs. AI-driven engineering

Slide 25

Slide 25 text

Don’t look down on vibe coding

Slide 26

Slide 26 text

Vibe coding is good for building prototypes and testing ideas quickly

Slide 27

Slide 27 text

AI-driven engineering is for building products

Slide 28

Slide 28 text

Vibe coding should be the starting point

Slide 29

Slide 29 text

Deployments and operations are what separate vibe coding and AI-driven engineering

Slide 30

Slide 30 text

People and processes (manual & automated) support actual products

Slide 31

Slide 31 text

Constant updates + QA (human & machine)

Slide 32

Slide 32 text

No content

Slide 33

Slide 33 text

Need to anchor to business goals

Slide 34

Slide 34 text

But the cost of experimentation is cheaper

Slide 35

Slide 35 text

Let’s talk about some basics

Slide 36

Slide 36 text

Step 0: spend some time building your skills for brand, colors, assets, etc.

Slide 37

Slide 37 text

Skills like Impeccable can show you what should be stripped to avoid “Claude-isms”

Slide 38

Slide 38 text

A basic vibe coding to AI-driven engineering workﬂow Start in Claude.ai / ChatGPT, etc. Create repo Vanilla prototype Claude.ai / ChatGPT to build md context Move to local filesystem init Enforce testing policy Configure deployment environment Automate deployments Deploy regularly

Slide 39

Slide 39 text

The “vanilla” prototype: ● HTML ● CSS ● Vanilla JS

Slide 40

Slide 40 text

Next step: make it work

Slide 41

Slide 41 text

Have the web agent (Claude.ai, ChatGPT) build a simple working version in your target framework (Rails, Next.js, etc.)

Slide 42

Slide 42 text

Once you have a working version that the stakeholders like, have your web agent build a full product spec with context

Slide 43

Slide 43 text

Place all of these ﬁles into a project folder on the development machine

Slide 44

Slide 44 text

Git init Create the repo on Github Connect the remote repo Claude / Codex init

Slide 45

Slide 45 text

Remember the diﬀerence between vibe coding and engineering? Deployment and operations!

Slide 46

Slide 46 text

Go in to planning mode, plan the roadmap. M0 should be the simplest possible version (“Hello world”)

Slide 47

Slide 47 text

Place product spec and other docs in /docs Place prototype ﬁles in /prototype Set rules in context ﬁles to ensure these never make it to public viewing

Slide 48

Slide 48 text

The key is to get the simplest version possible live in the production environment ASAP

Slide 49

Slide 49 text

Why? Because repeatable, automated deployments are a key factor in project success

Slide 50

Slide 50 text

Continuous Integration / Continuous Deployment

Slide 51

Slide 51 text

Just use Github Actions to build these

Slide 52

Slide 52 text

Create more rules about writing and running tests

Slide 53

Slide 53 text

Dev, staging, production

Slide 54

Slide 54 text

Dev, staging, production

Slide 55

Slide 55 text

Exploratory testing vs. regression testing

Slide 56

Slide 56 text

Unit tests / specs, integration, browser tests

Slide 57

Slide 57 text

The faster you can get through iterations, the more successful your products will be

Slide 58

Slide 58 text

We currently deploy 5-10 times a day during sprints

Slide 59

Slide 59 text

Things we don’t have time to cover today: Database selection Host selection Cloud vs. metal Object storage Logging Security & Audits Team Collaboration Context ﬁle management APIs, MCPs, and security for these And more…