Confidential · © 2026 Space-O Technologies Inc. · Prepared exclusively for DMA. Please do not copy, save or share.

Section 1 · Website AI Salesperson

1.4Objection handling

Built — needs DMA's data
No.
1.4
DMA's reference
Decision Book p.10
Phase
Phase 1
Autonomy
L4 · Autonomous
DMA's decision
Yes

What DMA asked for · Decision Book p.10

The chat knows how to answer the usual objections: "too expensive", "we already have an agency", "just researching", "need to think about it".

Why it matters to DMA: Objections are where most website chats die. Handling them well keeps the conversation going.

In our words, from DMA's Product Decision Book v3, page 10.

How we do it

  • Ten kinds of objection are defined: too expensive, already have an agency, just researching, need to think, no budget right now, needs someone else's approval, a bad past experience, wants guarantees, doing it in-house, and other.
  • Each has an approved answer in DMA's knowledge base, with guidance (for example: never criticise the other agency; offer a second opinion, don't push for a switch). The answers are versioned, and only approved, public ones can be used by the chat.
  • When a visitor objects, the strategist looks up the approved answer and gives it in a sentence or two, in its own words but only with the approved claims. If a sample case study matches the visitor's industry, it adds one line from it; if nothing matches, it doesn't mention case studies at all.
  • The objection is recorded in the chat's notes, so the salesperson sees it in the Nutshell summary before the call.
  • Today the answers are sample wording written for the POC. DMA's own answers, built from its real sales history as the book asks, replace them without any code change.

The flow

It plays on its own while it's on screen; hover or use the controls to pause or step through.

look uplook upmatchon bookingon bookingVisitor objects"We already have an agency"AI strategistspots the objectionApproved answers10 kinds, versionedCase studyonly if the industrymatchesClaims checkonly approved wordingReplyone or two sentencesChat notesobjection recordedNutshell summarythe rep sees it
Step 1 of 5

The visitor objects

A dental practice owner says they already work with an agency.

  • Person
  • AI
  • Stored data
  • Rule in code
  • Screen
  • Nutshell

Where the data goes

The same six stages on every page. Nothing reaches Nutshell except through the write gate.

  1. 1Website chat

    Source

    The objection, in the visitor's words.

  2. 2Sync

    Copies Nutshell changes into the bridge database

    Later: DMA's won and lost deals from Nutshell feed the answer library.

  3. 3Bridge database

    A copy of the CRM data, plus what the AI works out

    The approved answers and case studies, each versioned.

  4. 4AI

    Claude models, only through DMA's own gateway

    Finds the answer and a matching case study, and says it briefly.

  5. 5Write gate

    The only way back into Nutshell: checked, approved when needed, sent once

    Writes the chat summary, objection included, onto the lead.

  6. 6Nutshell

    The system of record

    The rep sees the objection before the call.

The tables behind it

Drawn from the POC's database catalogue: structure only, no data.

ai.kb_documentsPKdoc_keytextPKversionintegerkindtextaudiencetextclaimsjsonbstatustextIDapproval_iduuid+ 10 more columnstax.taxonomiesPKkeytextnametextapplies_totext[]active_versioninteger+ 5 more columnstax.nodesPKiduuidFKdimensiontextFKversionintegercodetextlabeltextdefinitiontext+ 7 more columnsconv.messagesPKiduuidFKconversation_iduuidseqintegerIDparticipant_iduuidsender_kindtextmetajsonbIDdecision_iduuidIDagent_version_iduuid+ 13 more columns

PK primary keyFK reference the database enforces (solid line)ID reference kept by id (dashed line)

TableWhat it holdsColumns
ai.kb_documentsDMA's approved content: services, FAQs, objection answers and case studies, each with its approved claims.17
tax.taxonomiesThe label sets the AI may use, such as the ten objection kinds.9
tax.nodesEach label in a set, e.g. "objection.already_have_agency".13
conv.messagesThe chat lines; each AI line records which approved sources it used.21
How the tables connect (7)
ColumnPoints toKept by
ai.kb_documents.approval_idctrl.approvalsThe application (by id)
tax.nodes.dimensiontax.taxonomy_versionsThe database (foreign key)
tax.nodes.versiontax.taxonomy_versionsThe database (foreign key)
conv.messages.conversation_idconv.conversationsThe database (foreign key)
conv.messages.participant_idconv.participantsThe application (by id)
conv.messages.decision_idctrl.decision_logThe application (by id)
conv.messages.agent_version_idagent.agent_versionsThe application (by id)

Worked example

Synthetic demo data: every name, business and number is made up.

A dental practice owner says they already have an agency (BrightSmile demo site)

1 / 5

What you see

1.The visitor gives some context first

"We're a dental practice with three locations." The strategist replies, and the note-taker records the industry and the number of locations.

A dental practice owner says they already have an agency: the approved answer, plus the matching sample case study.
Callout 1

What happens in the system

  • LLM gatewayfirst reply

    Claude Sonnet 5 through the AI gateway: 2,745 tokens in, 45 out · $0.0082 · 2.6 s.

  • LLM gatewaynote-taker

    Claude Haiku 4.5: 1,491 tokens in, 98 out · $0.0020 · 1.5 s. Records: dental, 3 locations.

Right-hand values come from a real run of the POC on September 25, 2026 (trace tr_muh5tbo52jgpqd0smgrf8, tr_muh5tbqv18g3rgfbfdhda…). Rows marked “sample” are illustrative.

The approved answers the chat uses today (sample wording written for the POC, to be replaced by DMA's own), and one worked objection.

Objection answers in the knowledge base

ObjectionApproved answer (sample wording)
We already have an agencyThat's common, and it's fine. We can take a quick look at what's working today and where there might be gaps, with no obligation. If your current agency is doing a good job, we'll tell you.
Just researchingMakes sense, and there's no pressure. I can look at your website and share three specific things worth fixing, or send a relevant case study.
Need to think about itOf course, take the time you need. Would a short summary of what we discussed help?
Doing it in-houseThat can work well with a strong internal team. Some teams use outside help just for specialist pieces or busy periods.
Too expensivePricing depends on what you need, so a strategist confirms it after a short conversation. Often the best start is the one or two things most likely to move results.

One objection, worked through

Visitor
"We already have an agency, so I'm not sure this is for us."
Recognised as
Already have an agency
Answer used
Objection: We already have an agency, version 1
Case study
Sample case study: a three-location dental groupUsed because the visitor runs a dental practice. A law firm would get no case study line.
Reply
"That's common, and it's fine: if your current agency is doing well, we'll say so. In a sample case study, a three-location dental group grew new-patient calls by 41% in nine months. Would a quick second opinion on your site be useful?"

In the running POC

A dental practice owner says they already have an agency: the approved answer, plus the matching sample case study.

1 / 2A dental practice owner says they already have an agency: the approved answer, plus the matching sample case study.

  1. 1The objection, in the visitor's words.
  2. 2The approved answer: no pressure and no criticism of the other agency.

Website chat · BrightSmile Dental demo site. Captured from the running POC on September 25, 2026; synthetic data.

See it live

Opens the exact screen in the running POC, in a new tab.

OpenWhereWhat to try
BrightSmile Dental demo siteDemo prospect websiteSay you run a dental practice, then that you already have an agency.
Knowledge baseDMA AI dashboardThe approved answers and case studies, with their versions.

Status

Built — needs DMA's data

Built — needs DMA's data. This capability runs in the POC today. It needs DMA's own content or data to go live; no code change is needed for that.

1.4 Objection handling · DMA AI proof